Google Cloud is 50% cheaper than AWS | Hacker News Reader

Google Cloud is 50% cheaper than AWS

thehftguy.wordpress.com

598 pointsyarapavan392 comments

Slightly unrelated, but since there seems to be some Google Cloud people in this thread: I see that Google Cloud account is linked to my Google Account. What happen to my Google Cloud instances if my Google Account got suspended by Google due to other reasons (e.g. Pixel stuff from the other day, or because of YouTube, or etc.) and I did not bought support tier?

On GCP, you create a separate Project entity that can have any number of Google Accounts linked to it. In that way, the actions of your personal account don't affect the GCP project (and vice versa).

If you're worried about access, you can establish a service account with Owner level access. Or you can add other Google Accounts to the project. (Here are the docs on that: https://cloud.google.com/iam/docs/overview)

I personally have all my Google stuff separated into 3 accounts (work, personal email, and personal Google-y things). My work account has access to the projects on GCP, along with some coworkers. That puts up enough of a firewall between various services so that if Google throws down the banhammer, it's not totally game over for me.

> On GCP, you create a separate Project entity that can have any number of Google Accounts linked to it. In that way, the actions of your personal account don't affect the GCP project (and vice versa).

My understanding from the recent Pixel-orders-account-banning incidents that Google had nuked even marginally related accounts, including gmail accounts whose only relation was being the recovery e-mail for one of the offending accounts. I just read this in one of the articles, so I don't know if it is true, but if so that seems to devalue the possible solution of just linking a number of different Google accounts to your Google Cloud account with owner access to the project. In fact that might be a larger concern if any one of those accounts raises the ire of whatever Google department does the nuking, they might extend the action to all accounts associated through the Google Cloud project. Hopefully not likely, but this stuff worries me.

Logical self-preservation in action.

And it is a telling indictment of the corporate character displayed by Google:

- Exploitable for minor personal convenience ("Use Gmail! Use Gcal! Simple and free!")

- Not to be trusted under any circumstances with important matters ("We can evaporate your existence from the internet for all practical purposes and there's not a damn thing you can do about it.")

Some day, it will become important for us to own our own lives again. And I say this as someone who runs his business on GApps. I'm as guilty as everyone else.

Curious what you mean by "own our own lives again".

Are you planning to run your own email servers (or any other service that Google provides)? Or just switch to another vendor?

> you can establish a service account with Owner level access

Never realized I could do this. Is my understanding correct that Google Cloud won't suspend a project as long as there's a service account associated with it?

I am compelled to create a HN account just to relate the below.

A year ago while planning to move from AWS to GCE, I created some instances to test under a company email id. Our emails were hosted under google apps. At some point, the email id got deleted for other reasons. I was in for a rude surprise. The instances disappeared, as I found out when the test urls didn't respond. Apparently, GCE deleted the instances without warning. I was later told by support staff that I should add another admin email to prevent mishaps like this. Seriously.

Anyway, we use GCE these days along with AWS and I quite like GCE. Of course, we now don't delete any email ids :-)

Google Cloud Support here. If you create Cloud project with a single owner (email), and that sole owner's account is deleted, the project is considered orphaned and will first be suspended, giving you some time to get touch with us if this was not intended, and then deleted after around 6 weeks. (Most running resources are stopped at suspension, but account data hangs around until deletion.)

I know this seems harsh, but there's really not a whole lot of other sensible options. The former owner can't be contacted, since the only email we have for them no longer works, and the project is literally unmaintainable, since nobody has admin rights to it anymore.

So TL;DR, for anything remotely important, please assign multiple owners and use role accounts (admin@corp) instead of users (bob@corp).

That is good to know, however my first reaction when reading this is we shouldn't be using G Suite for admin@corp or bob@corp, but rather if I am going to be seriously dependent on Google Cloud I should perhaps move e-mail for said @corp to Office365 (or whomever else actually does e-mail these days). I don't know what the solution/answer is here, but it worries me that actions (yes even bad spam-y TOS violating actions) that someone takes using our company's google accounts can ripple across the Google ecosystem and screw important parts of our company's infrastructure.

Why does Google not say "deleting this email will destroy these accounts attached to it"? Surely it can't be too hard to find which accounts an email is attached to, then check those accounts for other owners?

Why doesn't GCP force you to do this if it's best practice for anything remotely important? Surely your target customer generally is doing things that are quite important for them.

An email account could have ben recovered. Or you could have established identity of the user some other way, e.g. the credit card or the billing address.

I don't have any experience with GCE but I do have a lot of experience with AWS. What does an email id have to do with your cloud resources? "We now don't delete any email ids" implies you are worried about the same happening again. This makes me wonder why you don't confidently know what an ID is associated with so when you delete you are confident about not breaking anything else.

I think it could be viewed in a different light. It provides a real sense of control over resources (which should be well documented in the first place). If there are rogue resources and objects being utlized unauthorized in an organization it makes it short work to eliminate such things.

As someone whose google account is tied to a scary number of services, e.g. our company's G Suite, Fiber, Music, Play, Store, Cast/Home, Nest, Adwords, Analytics, Webmaster Tools, etc., and is in the process of migrating a bunch (but not by google standards) of data and instances from AWS to Google Cloud the thought that some screwup in some random service gets my account obliterated and my company loses access to our Google Cloud resources (because I'm the admin account) scares the !*^%# out of me.

Trust me. Single sign on is awesome. You WANT to have single account to manage all of that, and you should enable two factor authentication on it.

The alternative implying 100 separate accounts is so much worse.

It's like certificate authorities - it's awesome how easy it makes things if you can trust a single authority. Until the day you realize you can't actually trust it.

Completely disagree. I want a separate account for all my services. SPoF are a terrible idea. I miss when all my google stuff was separate accounts

Single sign on = single point of failure.

For corporate accounts, I want the various cloud services to use systems that have been around since forever and a day. I want to be able to use systems like Kerberos to give Google Apps a service ticket so I can manage passwords on my server, with google only getting what user has successfully logged in. The cloud has set us backwards in single sign on, since everyone wants to be the gatekeeper.

My Google account got suspended and I wasn't able to log into GCP. Instances kept running but obviously it's a lame duck when I had no management access.

Issue resolved with a call to a Googler friend :/

It seems 'having a Google friend to call' being the only way to talk to them comes up often.

I am a Google Cloud Platform PM.

Googlers are very active in many forums, including HN, Stack Overflow, mailing lists, and other forums, not just via paid support.

If a customer is having trouble reaching someone at Google regarding Google Cloud Platform issue, I'm sorry to hear that, and would love to hear what was tried and did not work so we can improve the process, but we are listening and paying attention, and happy to hear constructive criticism.

It's definitely NOT the intent that you must personally know someone at Google to resolve an issue with your Google Cloud Platform account.

I am a Google Cloud PM.

Sorry to hear you ran into this issue, and glad that it was resolved.

Just curious: did you try any channels of getting in touch with anyone at Google before reaching out to your friend? And if you did, and they did not work, we'd definitely like to know so that we can fix this.

We'd certainly like to make it easy for anyone to resolve issues with their GCP accounts easily, without requiring you to have a friend or acquaintance at Google to help fix this.

I know this is an afterthought, but you can create service accounts.

You should use separate accounts for a business. If you want to sell someday, you'll be glad that you don't need to go through your accounts and delete 35k personal emails.

Yes, I agree that I should have a separate account for business. But my concern was that Google has a lot of products and services that I'm not sure where I might tripped the trigger that got my account suspended.

For example, if my business email were hosted with G Suite, and my admin G Suite account got suspended for some random reason.

Making separate accounts might not be enough considering they allegedly banned accounts related to each others by recovery address. Why would you think they would not do the same with accounts sharing occasionally the same laptop, the same ip address, and the same first and last name ?

To my knowledge there is nothing preventing an automated banning machine to start firing at random on someone's accounts. And since, it appears, you have absolutely no recourse (except having friends working at google), and they are not held accountable for these "false positives"... You better start planning for it.

Can't you add another admin email, log in to that, and remove your own? I don't use GCE so I'm asking seriously.

You can actually do this by assigning another email as project owner.

A buyer might pay more if they were confident you included all relevant information, including communication history.

This is a great question. Common sense wise to me it's an issue. And given how large a company google is I don't think it even matters what the google cloud people who might respond say. It makes sense to simply isolate things like this.

Slightly related - had a AWS account, the account tied to it was for business use only - at some point I had credits/balance on the business account and someone gained access to the regular amazon account and bought some digital goods, so I notified amazon. They locked the account AND locked me out of AWS. This particular account was not important, but it took weeks to get access back.

I finally had to call my rep at AWS and said - you are putting your business with us in jeopardy (On other accounts that we spend significant money on) because its taken me so long to resolve this. Got it unlocked in 24 hours, but the whole thing left a sour taste in my mouth.

I suspect Google WILL have this issue as well, it has the potential to be a huge issue for your business.

Alternatively never had this issue when I've COLO'd - food for thought.

> Alternatively never had this issue when I've COLO'd

Taking it one step further in precaution (and from experience): We have some servers colo'd at a small datacenter. We use it strictly for our operating website and don't put anything for customers in that data center at all even though it would be convenient and save money. The reason is we don't want any bad actors that we might have as customers potentially impacting our main operations. This is in addition to vetting customers as well actually.

I don't think that should be a concern as both providers in theory can block or suspend your account. Maybe use a separate account for Google cloud for running world-changing-billion-user app.

That's why like most people publishing an app to the Play Store, I created a separate account just for this purpose. That way even if my account gets nuked for a reason or another, I won't lose half my digital life. same principle for GC.

Just remember not to put a gmail account as the recovery email for that app store account or you might still get nuked.

From what I understand, your google account wouldn't be suspended. It would be the individual services that could be suspended. Like you would still be able to access gmail if your YouTube account was suspended.

amen. thank you for raising this point: it is terrifying.

This. I don't want to end up pleading like a beggar to the google reps (or chase and plead to google employees on endless forums) to enable my sites (or worse, my clients' sites) which are disabled by google because their algorithm detected some "suspicious" activity on my gmail account.

With google, you have to be extremely cautious (no pun intended), especially after they have become big. AWS is far more reliable in this sense, and even Azure may seem better.

I'm curious if you speak from experience, or from the one article, where a customer let ports open, his VM was taken over by hackers, and proceeded to DDOS the world :)

"The numbers given in this article do not account for any AWS reservation."

While I agree that Google's pricing model is superior, the author's position on reserved instances accounts for ~40% of the cost difference.

In a drag race between instances AWS tends to lose. If you value the enormous feature and service surface-area that AWS provides it's a different story. Either way we win; both companies will engage in a brutal price and feature war for many years to come.

> While I agree that Google's pricing model is superior, the author's position on reserved instances accounts for ~40% of the cost difference.

Author here :D

Gotta take a stance to draw graphs, right? Comparing no reserved instance with the automatic discount is simple and realistic in my experience.

I could make pricing graphs accounting for 25% of AWS reservations and 75% of automatic google discount. In my experience, that's the proportions we _may_ have in practise. It would still be the same conclusion. Google is massively cheaper.

I could make TCO pricing graphs, accounting for the thing from the last paragraph, plus the human time (500$/day) required to understand, pick and manage reservations. I didn't do all the computations but so far it points towards a negative ROI (compared to not reserving), unless the infrastructure is big enough to have the economy of scale (hundreds thousands dollar per year). (Note: I really don't want to go against all the HN/reddit crowd who drank the reservations marketing koolaid, so purposefully avoid to talk about the TCO pricing model publicly ^^).

Thus, in all scenarios, google instances are massively cheaper. YMMV.

As a real life user of AWS I agree with this decision. RIs are not only a headache from a time perspective but we pay a non-trivial amount for a separate tool to manage purchasing them since AWS's built in tools aren't great.

Even with this tool we frequently make planning mistakes due to shifting business priorities or other reasons. Like - oh hay we have to upgrade our database instance but that needs a newer version because only the newer versions support something we need (i.e more than X GB ram). So engineering needs to modify the app to support the newer version.(Oh by the way AWS has a boatload of subtle restrictions like that.)

In this case we were sitting on a RI till EOY and for some reason you can't resell RDS RIs like normal ones (another subtle restriction) so we couldn't even partially recoup costs. Sometimes if you haggle with AWS Enterprise Support you can get them to help out if you're spending the money elsewhere but it's a pain. Good luck if you don't pay for that though (which itself is 15k a year or more).

What's the tool's name?

http://cloudcheckr.com/

https://www.cloudability.com/

And there is a 3rd competing tool that I can't seem to remember the name.

Probably CloudHealth

You really should have noted you were using on-demand pricing at the beginning of the article.

No one doing over $100k per month is matching utilization perfectly with RIs. Either you are not utilizing some percent of RIs in which case you are overpaying through breakage, or else you are paying on-demand rates for the instances which are not under RI.

The calculus involved with matching RIs to forecSted demand in a dynamic growing workload is NP complete. And any inefficiency delta pays AWS. There's a small part of the curve where your savings is minimal if you don't match perfectly, it it easily goes to negative on the RI change.

Otoh, nobody spending $100k a month is paying full sticker price.

Wait, $500/day to manage RIs? As someone who manages RIs for my org, that's way overblown. At most it'd be four figures per year. And once you get big enough, you leverage tools to help you figure out what RIs to get, which more than pay for themselves.

I would like to hear what platform bits of AWS that you find lacking in Google Cloud. Google's ecosystem of fully managed services is very rich broad and compelling, and in many ways far ahead of competition. (work on Google Cloud but don't get paid to post here)

> I would like to hear what platform bits of AWS that you find lacking in Google Cloud.

RDS instances with Postgresql.

This is the one that comes up. Hit me up offline. Also, you can always use any number of partners that provide this functionality.

Let me know if there are any others.

THis is in alpha on google

Thanks for the feedback!

I'm PM for another Google Cloud database (not Cloud SQL), but will share this with the team.

Elasticache (redis/memcache)

Postgres

Integrated domains (a la R53)

S-N/E/Q-S

For me, the most compelling attribute of AWS is familiarity. Google Cloud has traditionally been more forward with bespoke services that need to be adopted in order to use the platform effectively.

Familiarity with one way of doing things is a reasonable point. To assist here a little our SA's put together this:

https://cloud.google.com/docs/compare/aws/

Does Google have an equivalent for DynamoDB and Lambda? If so, I might consider switching.

Google has Bigtable and Datastore. Bigtable can be thought of as an events database, and Datastore is entities. Both very very compelling for DynamoDB users.

Google's Lambda equivalent is currently in Alpha. That said, there are bits like Cloud Logging and AppEngine that offer unique-to-Google functionality that's somewhat comparable.

(Work on Google Cloud)

DynamoDB and Datastore are mostly equivalent in terms of utility and pricing. GCloud also has Big Table if you're dealing with really large data sets/qps.

Lambda and Cloud Functions are competitors, though there are some things you could do with Lambda + API Gateway that you can't quite do with Cloud Functions just yet (pretty narrow scope, though). I prefer the environment for Cloud Functions, subjectively.

PAAS offerings for queue (SQS), notifications (SNS), email(SES) etc

Thanks - GCP doesn't match 1:1 on any of those. However, there are alternative (and for many use cases much better) ways on Google.

There's PubSub, Cloud Logging, Firebase (including Firebase Messaging), and all 3 don't really have a great direct AWS equivalent.

Take a look at the AWS/GCP equivalents at [1]. Mostly full coverage, although, again, some services work differently (usually better I hope!).

Email is indeed handled through either SendGrid or Google Apps [0]. Spam reasons I believe.

[0] https://cloud.google.com/compute/docs/tutorials/sending-mail...

[1] https://cloud.google.com/docs/compare/aws/

For me, the beta tag on things like App Engine Flexible instances is the biggest concern. We've had "fun" with Microsoft Azure services being less than fully baked, or Microsoft totally replacing services with different ones. In contrast, AWS has been very stable in this regard.

Yeah that's reasonable. At Google Cloud "Beta" means very far down the line of maturity. But YMMV and buyer beware when using "Beta" for production at any provider.

I have no idea how much a dataflow will cost to run. It tells me it's running on x vCPUs, but what will the bill actually be?

I am a Google Cloud PM, but not on Dataflow specifically.

Do you know how long the Dataflow processing will take? You may be able to estimate the length of the full-size process (assuming it's batch rather than streaming) by running on a small data sample to see how many vCPU hours it takes and estimate from there, assuming your data takes a uniform amount of time to process.

Since Dataflow is running code that you wrote, it doesn't know how complex your processing will be for any given value, so it's hard to give any reasonable estimate. In general, automatically predicting how long an arbitrary piece of code will take to run on a given input is even harder than the Halting problem (https://en.wikipedia.org/wiki/Halting_problem), which is already undecidable.

Details on Dataflow pricing: https://cloud.google.com/dataflow/pricing

Pricing calculator to estimate cost for Dataflow and other products: https://cloud.google.com/products/calculator/

Hope this helps!

"is very rich broad and compelling"

But does it support commas?

lol!

Try using https://www.cloudorado.com/ . It will calculate price for specific configuration for AWS, GCE, Azure and more. And you can choose if you want long-term commitment or not.

A nice attempt, but the comparison seems way off for anything beyond the lowest setting on the CPU slider. Choosing "1x" compares an EC2 c4.large instance type to others with a single vCPU (like a $6 DreamHost instance).

"large" is the smallest instance with fixed CPU allocation that AWS is providing these days.

I didn't see any mention of Google's preemptible instances.

I've used them, and they're much cheaper than ordinary instances. They happen to fit my use case, which is a simple copybigfile-simulate-writeresults done over hundreds of days of market data.

Preemption does happen, but it tends to happen very early on in a simulation. Also you don't get billed if it's in the first 10 minutes. For some reason GCE won't let you auto-restart that instance, but you can just write an API call that accomplishes the same.

There's also the added benefit that your instance will die after 24 hours, so you won't get billed for leaving something on by accident.

If you create an instance group of preemptible instances it will automatically try to fulfill your demand - no need to write a thing!

And it's amazing. Over 99% uptime for 1/4 the price.

One thing I forgot to mention is I've had a couple of "we will shut you down" emails.

Something about the automated abuse detection seems to not quite work, and so I've gotten emails saying the account is suspected of DDOSing other internet users. It's a bit unnerving, but they've been pretty prompt at getting back once an explanation was put forward.

To be fair, wouldn't you then need to compare to AWS spot instance pricing?

Nor any mention of AWS spot instances.

Here's a thought: Things often go wrong.

Would you rather have your mission critical apps relying on Google's customer service or on Amazon's customer service?

We pay 10+k/mo for our AWS support, and it has never paid off, even with multiple phantom errors and failures... the top level support supposedly being paged out of bed for me doesn't have answers. Always had to simply restore from backup without a good report to share with mgmt.

Agree. It takes an act of congress to get through to an actual AWS engineer. AWS status site is a joke with how degraded a service needs to be to register a change: https://status.aws.amazon.com/.

I've actually had good experiences with AWS support. Use the phone call option. Online replies tend to take over 24 hours.

We've actually had good luck with AWS Support and have a dedicated representative that is always there to push urgent issues through support.

That's deeply concerning. Can you pls share your aws account ID with me? If you're not comfortavle doing that, pls share email ID where I can contact you. Thanks.

You can email me anytime randhunt@amazon.com

I have gotten both stellar and not-so-stellar replies from AWS Support. They used to be slower than ideal but these days it's pretty good. We also have a great line of communication with someone who coordinates SAs to help clients (not sure what his title is) and he can get stuff escalated/sped up. For the most part, we haven't needed that despite his repeated offers to, except in one case, when he was quite insistent which it turned out was for good reason.

Either way I would greatly rather deal with AWS support than Google Support, which as far as I can tell is an oxymoron.

AWS support has been rarely worth the cost in my various stops. Even at the highest support levels. Slow responses, having to deal with people sticking to support scripts, etc.

Are you speaking from experience with Google Cloud Support specifically?

(Disclaimer: I work there.)

Probably speaking from general experience with Google customer support, which is generally opaque. The few times I have corresponded with Google customer support it's been bad.

This is a common sentiment, but I have to say that it's not in line with my experience as a Google Apps/G Suite business customer. If there is an issue, I simply call the phone number, get a competent support representative promptly, and if they can't resolve the issue on my first call, they will follow up and stay on the case. I haven't used GCP yet, so I don't know whether the support will be more like your experience or mine, but I'm confident that Google does have the ability to deliver good customer support.

I'm speaking from my experience of losing access to my YouTube account (same username as I have on this forum) due to having used my gmail address to sign up. Google later started linking everything to plus and somehow my gmail account spawned a new YouTube profile that clobbered my old one.

I spent many, many hours over many years trying to get access to my old account again and never did. Email is pretty sticky, so I'm stuck relying on Google for it, but after this experience I'm overwhelmingly unlikely to ever rely on Google for another service I care about.

I've also had some significant technical issues with other Google services where there is no support (e.g. the talk plugin for Firefox simply won't install and the troubleshooting page suggests using Chrome), but losing my YouTube account is something I'll never forget.

Every experience I've ever had with Amazon's customer service has been fantastic. I wouldn't want to be one of their suppliers but as a customer, they're great.

This.

I can confirm that first hand; We're just moving our whole stack from AWS to GC and right now we're running everything in parallel with roughly the same amount of resources. AWS monthly bill: ~$1000, GC monthly bill: ~$600

How many months will you need to run in GC to break even on the engineer time spent moving at $400/month savings?

I would say around 8, but then again, the goal was to use Kubernetes (which we do on GKE), so much of the time was spent configuring it, and we'd use at least some of that time even if we stayed on AWS's container service.

[EDIT] Plus I'm the technical founder, so enjoy this stuff, which means a lot of the work was done in my free time.

There is no such thing as "free time".

You're just missing it. Take a walk around the block some time. Time is more than free when it is your own, especially if you're learning something new or enjoying yourself.

I disagree. Taking a walk around the block is time spent doing something that could be spent otherwise. Everything in life has a cost associated with it. I am not saying a walk around the block or spending extra time at work are bad decision. My point is no time is free, you are always trading one thing for another. Going to the gym, you could be studying. Studying, you could be growing your business. Growing your business you could be spending time with family. Its all a balance in life but to claim your time is free is naive of reality.

edit: To those down voting you are welcome to present a different perspective. A down vote by itself doesn't mean much but I welcome criticism and feedback on my point of view.

If you always require a short-term win, you will lose in the long-term. When the company is 100x the size it is today, the savings are quite substantial.

Odds are low almost all startups will be 100x the size they are today. Most fail. Don't optimize prematurely.

True, but if you DO hit that magic growth factor, it'll probably be too late to make a change of such magnitude. Better to get optimization experimentation out of the way (relatively) early.

For nearly all businesses the cost of service hosting is going to be similar to the amount spent on snacks. This seems about as useful as optimizing the brand of cereal you buy.

That's a great problem to have.

Pro: Can't make hot cakes fast enough.

Con: I wish I could switch ovens; a white one matches my kitchen's decor.

Assuming a $400/month engineer, 1 month.

How many servers did you have to move?

I'd assume that you can count them on your fingers if the bill is only $1000/month. That's a "reasonably" small scale migration, if not locked in on AWS services.

Agreed, we have 10 instances on AWS, but it's hard to judge the complexity of the migration just based on that because these servers vary in size, requirements, and purpose.

But all in all, yes, it is a fairly small scale migration.

Generally speaking, the number of servers has very little to do with complexity of migration or existential threat from downtime, relative to company size.

Generally speaking, the numbers of server is a good indicator of the amount of services, which has much to do with the complexity of a migration.

Yep, we spend around $45-$50k/month and are quite happy with our support from AWS.

Just curious, what type of AWS instances you use? And also is it on-demand instance or reserved instances? I'm using AWS for couple of years, exploring other options.

On demand: m1.medium mostly for Rails and Node apps and db.m3.medium for RDS (Postgres).

Now on GCE, we use Kubernetes, which means the type of instances is rather insignificant as they can be instantly scaled up when more resources are needed for containers.

m1.medium? That's two generations ago. Just by switching to m3.mediums, you'd get 23% savings. If you could see fit to switch to t2.medium, you'd be saving 40%.

source: http://www.ec2instances.info/

thanks for the details.

Seriously this is worse than premature optimization, I like to call it developer frugalness optimization. A $600 savings is not even worth the effort.

For one month, likely. How many months are they going to be running? Will they scale up? When do you decide the investment is worthwhile?

They talked about moving to Kubernetes, so the AWS vs GCP part of this is much easier to deal with.

Agreed - primary motivation weren't savings, but much smoother UX with Kubernetes on Google. We're adopting container-based approach to development (and DevOps), so we explored alternatives, and chose Google Cloud.

Thanks for sharing your experience. Why, acc to you, is Kubernetes on AWS a non-starter? Is it too difficult to setup? Or maintain? Or simply isn't the first class citizen, like it's on GCP?

Or does it just make sense to stick with GCP since K8s has Google's blessing?

Or...

Two reasons really, the first is that AWS has become this bloated mess of "stuff", making it increasingly harder to use, because so much time is needed to either constantly look at it to see what's new/changed or vigilantly document everything. GKE, based on my experience, uses more "convention over configuration" approach, so the UI is much less cluttered, easier to understand, and comes with good defaults out of the box. And yes, K8s is a first class citizen on GKE and it overall fits nicely in the google compute engine environment.

The only downside is that Google Cloud doesn't have a hosted DB offering for PostgreSQL, like AWS RDS, so it took me a while to set up everything properly.

Finally, from a purely subjective point of view, Google's Material design is easy on the eyes.

This is roughly what we ended up with for our stack: https://cl.ly/1z141g0e1w38 (The top three instances are K8S, then two GlusterFS instances which hold persistent volumes for pods and finally three PSQL instances that also run Redis Sentinels with a quorum of 2 - Redis itself is on Kubernetes as a DaemonSet)

why is anyone downvoting this?, comment author has a point.

Now, If only I could use it in the UK.

Seriously. Every time I sign up, I have no option to sign up as a individual, only company.

Friends in other regions of the world can sign up as individual, but it appears google ( in the UK/EU ? ) as decided for whatever reason ( I assume VAT calculation ) they won't offer it.

I can sign up to and use AWS ( as much as I really don't really want to ) as myself. Yet there are all of these really nice things coming out of GCP that I can't use because I simply can't enable billing. (that being said, I do use google app engine for my blog, and it's fantastic)

If any cloud project was a single dimension problem. For me it does not matter how much cheaper I could run an instance in GCP simply because we rely on services from AWS that do not exist in GCP yet. Another issue for me with Google is how they handle account problems. I do not want to find out the hard way that I should have done some extra safety measures just to keep our entire production environment safe. With AWS I do not need to worry, their customer first approach proved to be extremely useful over the years, and they were very patient with us even when we did something that was against their policy. My customer trust is not something that is up for sale and I have been disappointed with Google's customer support several times, this is simple cannot happen in a cloud infrastructure situation. I like GCP and Azure because they make sure that I get a good deal on AWS.

This guy has some other really great articles. My favorite is: "GCE vs AWS in 2016: Why you should never use AWS" [1].

[1] https://thehftguy.wordpress.com/2016/06/15/gce-vs-aws-in-201...

Does someone know why Google Cloud can afford to be so much cheaper compared to AWS? Are the costs to Amazon for AWS so much higher than the costs to Google for Google Cloud? Why?

AWS is actually a pretty high margin business (~25% operating margin is public, GM is not publicly known but good guesswork puts it ~50%). So, there's some hard data showing a lot of room for price competition.

If you'll allow me to get a bit more wibbly: I've been an engineer at both companies, and from a technical perspective, Google is much better at this stuff. In order to provide their own services, they've been operating at much larger scale for much longer than anyone else. There's a pretty long record of the new hot thing in the industry being something Google's been doing on the order of a decade. The basic deficit in experience is exacerbated by Amazon's relatively poor retention of its best talent.

Putting all of this together, it is my expectation that Google be able to build better services faster, and to operate them more efficiently. AWS started with a massive first mover advantage, but my expectation is that in the long run Google will have little difficulty closing that gap and pulling ahead, should it wish to.

> AWS is actually a pretty high margin business

Correct, see how much profit AWS contributed to Amazon in the past quarters.

> I've been an engineer at both companies

Same here, so I offer my perspective.

> Google is much better at this stuff

Not much, but noticeably better across the board. Overall it's substantially better when combine everything together. But as others point out, the technical superiority does not account 100% of the quality of product offering. Being in 2 companies, I think you have similar experiences. Otherwise, GCP should already eat AWS in 2012.

> Google will have little difficulty closing that gap and pulling ahead

I think so. But my expectation is that Google needs operate Cloud as its top priority, and treat AWS as a tougher competitor than MSFT in the search/browser war. The thing is, that Cloud is a much much bigger market than internet ads. The race will be long and difficult to predict. Amazon is a much more formidable competitor than MSFT. They have proven to be the best in retaining customer and build a brand. They are also extremely fast in execution. And they are not afraid to try crazy ideas and abandon them if proven wrong.

Plus they have a real customer service culture, something Google seems allergic to.

Edit: posts further down from mine seem to suggest my sentiment is outdated and Google takes cloud support very seriously in their teirs. https://cloud.google.com/support/

>Amazon is a much more formidable competitor than MSFT.

I don't know about that. Going further, majority of the spend is going to come from established businesses. They need software and not just infrastructure. And msft has those accounts. Has the reseller network. Has the software. I really can't think of any good b2b/productivity software that google wrote. Probably they haven't tried, but looking at what they announced last - data studio, survey and optimize - the talk was like they were super serious about it. The result is very, very far from the competition.

>Google is much better at this stuff.

Also worked at both companies. Google is not better. Google is a retirement home for engineers. Few creative things happen there (anymore--wasn't always true). If GCP is so star spangled awesome why is it still so far behind AWS? Why did AWS start literally 10 years before GCP?

Google reacts to competition while Amazon proactively responds to customers -- this, as always, is the difference that allows amazon to succeed.

> why is it still so far behind AWS? Why did AWS start literally 10 years before GCP

That seems to answer the first question. AWS had a multi-year head start, but GCP is catching up and has the fundamentals covered, and the primitives they offer are very strong.

You are currently employed by AWS, but I don't see a history of you working at Google.

A "bias" disclosure would be helpful here.

(work at Google Cloud and don't get paid to post on HN)

It would be desirable that Google has a decent margin in its GCE operations, otherwise there is a high risk that they cut it back or even shut it down.

I know Google and a couple other companies are designing their own hardware to reduce overheads like power and square footage. Does Amazon do the same?

AWS has publicly stated that they build their own hardware for a few years now.

One of the things I've heard (from googlers) is that by using KVM as the underlying hypervisor, they do live migration much better than Amazon. Using that, they can live migrate applications to the same servers and bin-pack much better. This allows them to squeeze more jobs onto the same hardware by leveraging their already dominant lead in machine learning to predict workloads.

Basically, they can do a better job than Amazon predicting current and future workloads. As a result, they can more efficiently place workloads and literally do more with equal or less hardware. That and the fact that they have been building their own hardware better means their hardware is also very likely more advanced. Those together mean they can do the same thing for cheaper and still turn a profit.

Amazon optimizes for doing something first and doing the most. Google optimizes for doing things smarter and as a result often does them cheaper. However, it also means that Google provides massively fewer services than Amazon.

> Does someone know why Google Cloud can afford to be so much cheaper compared to AWS?

They have better hardware utilisation and a lot of their infrastructure was fully-depreciated before they opened GCP.

Unlike AWS, Google couldn't design systems assuming a customer would bear the price of any moderate inefficiencies. So they've had more design pressure on cost from the beginning.

I also suspect their cost of networking is much lower -- they own their own fibre network and a lot of it will also be fully-depreciated.

I'd suspect that AWS has a pretty healthy margin on their instances, and Google might be willing to accept a much lower margin to gain marketshare.

The same reason that Google could launch Gmail with 100x as much free storage as Hotmail: Google's physical infrastructure is dramatically less expensive than everyone else's.

IIRC I think I have seen some people here argue it is because they have close to no customer support.

From Amazon however I have even heard about them forgive bills where the customer was really to blame.

All this is hearsay though and I don't think Amazon does this out of the goodness of Jeffs heart but as long as enough people think this is how it works it might be part of why people will use AWS even if it is more expensive.

Our support folks can comment here, but I would disagree with that point. Our support org is massive, there are varying tiers levels of Support SCE/CRE/SRE/SWE-level of support, all the way up to [0].

I'll also comment that Google's Eng/PM org is very engaged. It's not unusual for an engineer in charge of a service to help resolve a customer an issue.

[0] https://cloudplatform.googleblog.com/2016/10/introducing-a-n...

(work on Google Cloud but do not get paid to post here)

> It's not unusual for an engineer in charge of a service to help resolve a customer an issue.

This is not related to Google Cloud, but to GSuite: Or, if you get an engineer on the phone, through the support number, they just insult you, and hang up on you, as has happened to me once (back when the free tier of GSuite was still a thing).

> It's not unusual for an engineer in charge of a service to help resolve a customer an issue.

This is the norm with AWS, in case you didn't know.

Like someone pointed out, having a deeply engrained customer-focussed culture is different from having a massive support org that quantifies customer happiness as datapoints on a review board somewhere in cozy SV offices.

I am not sure that you can really use the size of your support org when talking about customer experience. Why don't you come up with a 3rd party customer survey about how your customers are happy with you compare to Amazon? If you are absolutely certain that your support is great it would prove the point.

> they have close to no customer support.

In the last three days I have been able, in my day job, to personally ask a Google PM to pull Google Cloud engineers into a shared slack channel to help.

It took us longer to set up the slack permissions for the googlers to join than it took for Google folk to identify who to pull in.

You failed to mention you are a major partner of Google Cloud. So the average customer probably should not expect the same response.

Well, AWS has only billing and account related support for free. Any other support is, of course, paid. And it'd be best to avoid commenting on their free support.

> IIRC I think I have seen some people here argue it is because they have close to no customer support.

Assuming that this is true, having no customer support is still better than the AWS support ;)

I'll toss my personal experiences in as well: we pay Google for support at the 'Gold' level[0] and we've been satisfied with the support they've provided. We've only had a few issues but they've gotten back to us quickly and it's been easy to access their technical staff.

[0] https://cloud.google.com/support/

My guess is economics are trying to steal AWS business. Low pricing will draw a % of AWS customers over to GCE and the prices will balance out. Frankly I am more interested in Google's support quality compared to AWS. I don't have any experience with Google but I know Amazon is founded on the premise of customer experience which carries into their support. How is google's support?

I am not familiar with GCP support but with their product support. Once I bought a device from Google and it broke on the first day. The device was completely dead, did not even turn on. When I engaged their support and explained the situation to the customer representative she kept asking me which light is on even though I explained to her 10 times that the device is completely dead. I was lucky because I bought the device through Amazon. I visited the website clicked on the dead device needs to be replaced button and in 10 minutes the replacement was on the way to my home.

I do not want to draw any conclusion here but when it comes to which company gets my money decision I always prefer Amazon.

The following applies to customer trust too:

"It takes a lifetime to build a good reputation, but you can lose it in a minute."

― Will Rogers

AWS is expensive as fuck. If i move the structure of one of my websites to AWS i would get my 99.99% uptime instead of boring 99% but it would cost avout $400 dollars instead of about $50. But sure thats missing the fact that i hafs to self manage right now.

It's Google using their typical strategy: spend lots of advertising cash to develop and then provide something for free / less than competitors, with the goal of pulling you into their ecosystem.

Under the Obama administration we didn't see any antitrust/antidumping actions because of how much Google managed to embed themselves into the administration. Don't know if it will change.

EDIT: fixed a typo

Your latter point is not really a point at all. There is vendor lock in at every corner with AWS. The big 3 all do this.

In my view, there's really only two things that can reduce cost enough to make a difference: 1. Reducing energy consumption, 2. Reducing drive / component failure.

I believe Google may be better at both of those than Amazon.

I've worked with both AWS and GCP (my day job is working on Cloud Foundry).

In every area that I've looked, GCP is so much better than AWS that it's offensive, bordering on obscene.

It's faster, cheaper, more configurable. The documentation is actually comprehensible. The APIs are consistent. There's a single console, rather than a fruit salad of conceptually different consoles under a single domain. The console is, in fact, navigable without invoking curses. Not even a small one.

AWS has the massive advantage of inertia. If you're deeply woven into the higher-level AWS services, I'd probably stay put.

But if you aren't, or you're just beginning, then holy moly you owe it to yourself to look.

In every area that I've looked, GCP is so much better than AWS that it's offensive, bordering on obscene

GCP's major weakness for me is locations. They will (if they deliver on their plans) get a proper global footprint in 2017, but right now it's a US and EU centric service that simply can't match AWS's global reach.

> In every area that I've looked, GCP is so much better than AWS that it's offensive, bordering on obscene.

I know. That's the feeling I have every time I compare them.

It's just borderline insane. There is enough materials to write an article: "100 little details that AWS got wrong but Google Cloud got right [and the 4 exceptions that are the other way around]"

That sounds like it'd be an amazingly useful article to read if you ever write it :)

>It's faster, cheaper, more configurable. The documentation is actually comprehensible.

The docs have more 404s than trumps twitter account... who payed/paid you to write this?

I haven't hit any yet.

The AWS docs are enormous, cover every angle, and take me quite a while to work out. I'm not very bright.

That said, Google's concepts are confusing in places. The one that caused me a lot of head scratching was the difference between forwarding rules and global forwarding rules. AWS has better diagrams in a number of places.

Thanks for the follow up and sorry for initial attack.

As I said below AWS docs could be dramatically improved for accessibility.

I don't mind. You're passionate about your work. That's a good thing.

>>It's faster, cheaper, more configurable. The documentation is actually comprehensible.

>The docs have more 404s than trumps twitter account... who payed/paid you to write this?

(ranman works at AWS..and talking to an AWS premier partner... the delicious irony)

Jacques doesn't speak for all of Pivotal. Its of course his personal opinion that GCP is way ahead of AWS. And Jacques, personally, is entitled to his opinion. There might even be merits to his statement here but he hasn't backed it up at all with data.

Pls do not pass off his statement here as an official stand by Pivotal. Unless of course that's what Jacques here meant to do (I wouldn't know)...

It is correct that unless I explicitly say otherwise, I do not officially speak for Pivotal. It'd basically be a terrible idea to give me an official mouthpiece, given how bombastic I am.

And, of course, my opinion of GCP is just, like, my opinion, man.

In my personal projects I have stuff on AWS. In fact I'll be adding more soon, since Pivotal Web Services is housed on AWS. And there's still a lot that AWS can do that GCP can't. It's just that, for the tiny slice of the world I've seen, I like GCP a lot more.

No matter which platform you prefer, everyone will benefit from the fierce triangular tug-o-war between Amazon, Microsoft and Google.

Especially if they're on a relocatable platform like Cloud Foundry (disclosure: Pivotal sells a version of this) or OpenShift.

Great job addressing those 404s.

Again, a weird thing for an AWS employee to say about an opinion from an AWS customer and an AWS platinum partner, and then accuse them of expressing those opinions due to monetary compensation.

In any comparison on "cheapest way to run a linux box somewhere" I would expect AWS to lose. AWS is a platform, not a server rental service. You start to see value when you build to the platform.

(work on Google Cloud but don't get paid to post here)

For me the two big things missing from Google Cloud are a PostgreSQL service and a Redis service. I'd also need an alternative to CodeDeploy, but I think I could something that wouldn't be too terrible to self-host.

> in many ways far ahead of competition

Yep. In my experience, when Google Cloud does something, they do it right.

Thanks for your feedback!

Managed Postgres.

Hit me up offline - this is one gap Google's had, I admit. On the other hand, you've always been able to run managed postgres through a partner.

Let me know if there are any others that you feel strongly about?

Direct Connect

Cloud Interconnect is the equivalent to this on GCP.

Even when the comparison is with a web-hosting provider such as GoDaddy or Digital Ocean? I want to host a few custom developed web apps on a VPS and I'm confused between AWS and DO/GoDaddy. Which one would you suggest?

Web apps on VPS sounds like something fit for micro or tiny instances, in which case a month pre-commitment might actually be cheaper if you look at OVH's Kimsufi programme: https://www.kimsufi.com/en/vps-ssd.xml. These are cheap, but in my experience reliable (after multiple suboptimal experiences with DO) KVMs.

While KS is probably more important for EU customers, the KS-VPS line has three regions to choose from: Gravelines (Northern France), Strasbourg (Eastern France) or Beauharnois (North-America). You can also DIY your own cloud with https://www.ovh.ie/vps/vps-cloud.xml, though I've never tested this.

Personal point: I may be very old-fashioned, but after using GCE and AWS for various tasks, I still prefer the transparency of pre-committed billing. E.g., I don't know if this changed by now, but I ran into dozens of inexplicable administration errors with GCE, to the point where I couldn't manage my own projects anymore. Plus, after creating billing details, I couldn't detach my credit card from the platform. The only way around that was to disable all projects, but that still left me with an uneasy feeling.

I don't have automatic billing setup for KS/OHV. I get billing reminders 30/15/7 days before each cycle and authorize each payment.

DO is more comparable to AWS than GoDaddy.

Essentially a VPS provider is someone who rents you a server instance on a much larger server running other instances of other people's servers. You are in complete control of your server instance. You have to manage all the application installations, firewalls, security, etc.

A GoDaddy or Bluehost is someone who gives you folder space on a server. They have the web server, database, and any other application all setup for you already. Typically those types of providers only allow you to serve static content, or certain scripting languages (mostly PHP).

It depends what stack you wrote your web apps with for which provider you should go with. With a VPS, Linode is going to be cheaper than Digital Ocean, which in turn is cheaper than AWS when you compare specs to specs. If you're app is a simple PHP app that talks to a database, you could probably host it way cheaper on GoDaddy or Bluehost. You just wouldn't have very much control over all the different versions of the DB/PHP, etc. that you might want.

I would suggest a DO droplet since the learning curve is very small compared to the AWS services, the cost is predictable and they have a decent bandwidth allowance.

If you want to compare for prices, just search for "VPS with hourly billing" and you should find some alternatives to DO which may be worth considering.

And if you want to dig deeper, I have written a book [1] which I hope can clear the confusion and give you a better idea of what hosting you need for your apps.

[1]: https://www.hostingforappdevelopers.com/

This. Some of my clients are on one of the spectrum with lots of PaaS offering in use especially features such as queues. Others prefer to be agnostic and won't use any specific features. I guess it depends on what you want and how comfortable you are administering Rabbit etc.

We use many services on AWS (S3, ELB, RDS, RedShift...). The costs of the EC2 instances is > 70% of the total bill. It's in the same order for the other deployments I've seen.

The thing is, the AWS pricing strategy is global and most services are built on top of instances.

I hate it when the evangelist swarm the hackernews comments.

On another note, I'm not surprised Google cloud is cheaper as its trailing behind and offer no other advantage but price to catch up. GCE, Azure and AWS all pretty much match each other technically, so I suspect the one with least customers to always offer the better value. So if GCE were to ever become number 1, I'd suspect it's price to rise and others prices to lower.

"The numbers given in this article do not account for any AWS reservation"

Well that's a huge caveat. Anyone intending to be a serious user of AWS will use reservations. The discount is large, well over 50% off IIRC. And then there's the spot instances/spot fleets with even steeper discounts.

For better or worse, AWS's pricing is more complicated then its competitors. A serious comparison would have to include those details.

Actually, because you've got to pay in advance a certain amount to get the reserved price, we calculated at my last job that buying on-demand instances was actually cheaper than paying the reserved instance pricing listed, because of the rate at which AWS drops the price regularly. Meaning, if you compare the reserved price with the amount you would pay for on-demand instances and subtract the amount you save when Amazon drops the price (which they do pretty regularly, and have in the past), the on-demand price actually ends up being cheaper in the long run.

I think you meant "on-demand", not "spot". Spot is hands down cheaper. Reserved saves you 30-75% depending on instance type and duration. Spot instances save 85-90%.

As for on-demand, AFAIK the only time a discount came close to 30%/year was in 2014 when Google Cloud slashed their prices. So...yeah, it might be possible, but I doubt it.

Hah, I don't know how you were able to see my first edit. It was up for less than ten seconds and I didn't have your comment for some time after I edited it!

"Anyone intending to be a serious user of AWS will use reservations."

Not in my experience... It's impossible to predict the kind of usage you're going to need over the next 1-3 years, especially when you're first getting started with AWS. If you underestimate, you waste money. If you overestimate, you waste even more money. Google's concept of "sustained use" discounts is _much_ more appealing to me.

> when you're first getting started with AWS

Then I respectfully submit that you are not (yet) a serious user of AWS.

> It's impossible to predict the kind of usage you're going to need over the next 1-3 years

I call BS. Office 365, JIRA, Slack, Loggly...these all have monthly price plans, and discounted yearly plans. People managed to predict needs, despite a uncertain future of their needs. In the case they underestimate, they just purchase more at anytime. In case they overestimate...I don't know; I just know that a price comparison would very likely use JIRA's yearly prices, not its monthly ones.

Unlike those examples, reserved instances can be bought and sold on the AWS marketplace (https://aws.amazon.com/ec2/purchasing-options/reserved-insta...).

I understand (and agree) if people criticize the complexity of AWS EC2 pricing. But if you're not going to use reserved instances for your non-trivial stuff...you're flushing money; go with another provider already.

> I call BS. Office 365, JIRA, Slack, Loggly...these all have monthly price plans, and discounted yearly plans.

These services have different, not comparable pricing models. See http://www.joelonsoftware.com/articles/CamelsandRubberDuckie...

They only sell packs (like 5, 50, or 500 users). All customers have to pay for the superior pack. There is little price prediction, you just _seriously_ overallocate to the next plan available. That's a revenue optimization strategy for the company. It's very costly for the customers.

> Unlike those examples, reserved instances can be bought and sold on the AWS marketplace (https://aws.amazon.com/ec2/purchasing-options/reserved-insta...).

Nope they cannot... UNLESS you're American, and you have a bank account in America and in dollars, and you have an official American tax number, and you fill some paperwork... and more unknown hiccups you'll only find out too late! (I couldn't continue the procedure to see what's next, didn't fit the criteria at this point ^^).

well you just buy the lowest possible and start new instances and stop them when needed, when the lowest possible raises you just reserve more.

This article is interesting and the author has some good points about the lack of small non-burstable instances in AWS. There are plenty of things to gripe about.

But the headline assertion is ultimately unsupported except in terms of simple comparison graphs of paper numbers about the raw CPU and memory numbers. Are the CPU units comparable? Is the networking what it's cracked up to be? How easy is it to autoscale? Is the capacity you need available when you need it? How long do instances take to start? What are the dynamic storage options? How does disk IO performance compare?

I'd be really interested to read an article that attempted to break these down and do a real comparison. But this article doesn't even attempt a real world comparison.

> Are the CPU units comparable? Is the networking what it's cracked up to be? How easy is it to autoscale? Is the capacity you need available when you need it? How long do instances take to start? What are the dynamic storage options? How does disk IO performance compare?

In order.

Yes. Google networking is superior (cheaper & faster). Variable, depends on your application/workload, not just the cloud. Yes. < 30 seconds on Google, 1-3 minutes on AWS. lcoal SSD, remote SSD, or remote HDD on Google VS a mess of many complex & expensive disks on AWS (it would take more than a blog post to explain their disk offerings). Google disks are 3-10 times IOPS and/or bandwidth, lower latency.

This article is just on basic pricing. I plan more articles for the future. Starting with one on disk benchmark and one on network benchmark.

Does anyone know if there are any plans to support Postgres by the Cloud SQL? Because currently it's the major deal breaker to me. I have a SaaS business that works on Postgres and we don't plan to replace it with any major SQL alternative.

There's some indications that Google would support Postgres as part of the CLoudSQL in the future. But there are already multiple DB-as-a-Service offerings that provide PostgreSQL in Google Cloud. My company, Aiven (https://aiven.io) is one of the providers and I believe Compose, DatabaseLabs and ElephantSQL also provide managed PG in Google Cloud.

Thanks for the recommendation, we will give it a look for sure.

Lots of discussion here too: Which cloud provider to use in 2016? AWS or GCE? [0]

[0] https://news.ycombinator.com/item?id=11515505

I want to like Google Cloud, but it just seriously lags AWS.

Amazon got everything so right with an ID and secret. GC's oauth is just cumbersome, and the CLI tools are nowhere near as user friendly as aws-cli.

Two other recent examples - the lack of granularity and inflexibility of GC's IAM perms (you need to faff around with roles which even as an Owner you can't create - they need to be done at the org level, WTF?), and GC support is miles behind.

I opened a support ticket for a critical issue recently, and the initial response I received when someone looked at it fell into the repeating-back-to-me-what-I'd-told-them-in-the-ticket category. Then they suggested I "check the permissions are right", which as I told them "I don't know what's wrong or what needs checking, that's why I opened the ticket". It's just lucky the issue didn't affect our prod account or we'd have been in real trouble since they're still investigating even though we're probably one of their top-tier customers.

AWS support has always been second to none for me. GC support has always for me been second to, well, Amazon...

I actually like the gcloud CLI -- I wish its output was more tunable (CSV, Table, JSON)... I also think it's weird they use a ~/.boto config.

Have you tried '--format'?

that was easier than I expected.

I recently visited the google cloud pricing page and had a hard time making any sense of it all. The pricing isn't at all clear. DigitalOcean on the other hand, you take one glance at their pricing page and know exactly what your getting. Google can learning something from that.

Thank god I'm not the only one. I keep hearing about how opaque AWS pricing is compared to GCP, yet I personally cannot make any sense of google's pricing. It's difficult to explain, but I struggle ti figure out how much I'll be paying over, say, a month for a GCE instance. It seems like there are too many variables to keep in my head.

But I'm just a tinkerer/hobbyist. So maybe the reason is because I don't have a clear idea of what my monthly utilisation is going to look like, or whether it's going to be consistent. I imagine this would be much clearer for those running businesses. I dunno...

GCP pricing is far simpler than AWS - just visit any of the product pages to see the unit costs.

And they have a calculator: https://cloud.google.com/products/calculator/

If you're talking about Compute Engine utilization for billing discounts, they just mean how long you have it running continuously over the course of a month.

How much do google use their own cloud internally? I'm curious and would better appreciate Google Cloud if I could see it being dogfooded with something substantial.

My understanding is GKE on GCP is kinda sorta a rewrite of the Borg system Google has used internally for years. Internal services at Google are encouraged to use GKE, but there's little incentive to migrate from the Borg infrastructure that has a long track record for reliability. So currently only a marginal amount of Google services use GKE, but as old services are retired and new ones developed, the share of services on GKE should grow steadily over time.

In July, Anvato joined Google. They're using Google Cloud to power video processing solutions for customers like NBCUniversal and Univision.

http://www.anvato.com/anvato-joins-google/

A bit off-topic, but

- AWS has a 12 months free tier [1]

- Google Cloud has a 2 months free trial [2]

That will make a huge difference when I'll have to choose.

[1] https://aws.amazon.com/free

[2] https://cloud.google.com/free-trial

Great point. I think it's a matter of different philosophies, and both are great for customers!

- AWS gives you specific usage allotments per-month for specific services.

- Google gives you $300 cash for 2 months to use however you please. just don't mine bitcoins or generate email spam :)

- Some folks did the math on AWS free tier to total $247.20 [0]. That is, if you 100% utilize all the allotments.

- Google BigQuery, Firebase, and AppEngine also have perpetual free tiers.

[0] http://searchcloudcomputing.techtarget.com/tip/How-much-are-...

(work on Google Cloud)

Thank you for the clarification. :)

Talking of EC2, AWS free tier only allows t2.micro instances (1 vCPU, 1 GiB RAM). Might not serve any use to many.

It seems like Google Cloud free trial in several countries (at least in LV,LT,EE) is only available to businesses.

I wanted to give the free trial a go, but wasn't able to register because it did not let me choose account type as "individual".

Comparing only instance prices is not fair, imho. Amazon charges that because all of the services/ecosystem they have. Google Cloud is stepping up their game though, leading to cheaper prices in AWS.

Amazon has a very broad offering of great fully managed services, but Google's no slouch. In fact, many of Google's offerings can be very compelling - BigQuery, Bigtable, CloudML, PubSub, GKE to name a few.

Just in the "only instances" space you mention Google has a different strategy. A couple things to mention:

- Google doesn't have a broad spectrum of instance types. No storage-optimized or networking optimized. Instead, any instance can just get great storage attached to it, and any instance gets best-in-class networking.

- Google's Preemptible VMs are like Spot instances but fixed 75-80% off. Again, with less fragmentation against instance types + fixed cost, much easier to rely on Preemptible VMs imo.

- Google's Load Balancer is global, scalable, anycast IP driven, and backed by Google Network. If your packets originate in, say, New Zealand, they'll be talking to a "GCLB instance in our pop in Sydney", which will carry packets on Google's backbone to the VMs.

- Custom VM sizes - you can set your own VM/RAM combination for instances.

- Live Migration. Google manages instance health and maintenance for you, without forcing restarts.

(Work at Google but not on GCE.. and I don't get paid to post here)

I'm not sure I'd agree that Google has a broad spectrum of instance types. You can boost up individual components, but for example if I wanted an EC2 instance with as much local SSD space as I can get I could have 6.4 terabytes, while on GCE I could have 3 TB. If I want memory, Google's willing to give me 200GB, Amazon offers 10x as much.

My impression is that Google has a first class general purpose instance but you don't really get the breadth of options that EC2 will give you.

You bring up a good point. Amazon does give you a better "vertical scaling" story. I'll still challenge you on the "breadth" when it comes to EC2 - the philosophy is just very different. Why do you need a "IO optimized instance" if you want just fast disk - that notion just seems very foreign and arbitrarily-constrained on Google Cloud.

You bring up Local SSD. Google's Local SSD is just badass by comparison:

- 680,000 Read and 360,000 Write IOPS included in the cost [0]

- $0.218 per GB per month. Instance cost is separate.

- Again, you can attach these to any instance type (hence the point on fragmentation of instances on EC2)

- AWS goes up to 365,000 Read and 315,000 "First Write" IOPS. Only if you buy an i2.8xlarge [2]

- An i2.8xlarge is $6.82 per hour.

You do the math :)

And someone else did more comparisons here [1]

[0] https://cloud.google.com/compute/docs/disks/performance

[1] https://medium.com/google-cloud/new-google-cloud-ssds-have-a...

[2] http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/i2-instan...

The real problem I have is the low network performance. Yes, yes, before everyone jumps all over me and points to Jupiter etc.. I understand the problems in Pb/s bisection bandwidth for the large datacenters. That doesn't change the fact that I don't need an entire datacenter worth of stuff.. but I do need an Amdahl-balanced cluster. So big machines with wimpy (20Gb non-RDMA) networks prevent me doing my HPCish workloads on GCE.

Followed by waiting on GPUs and other user accessible accelerators of course.

There are only two notable differences:

1) The highest memory instance on Google is about 200GB, on AWS it's 2TB.

2) Google doesn't have GPU instances yet. They're been announced for 2017.

Other than that, google instances have more options and more flexibility.

Other than the fact that AWS has more options and flexibility google instances have more options and flexibility... hmmm

AWS spot instances are a fraction of the price, like 20-30%. In my experience, building the same processing cluster from AWS costs half as much as Google Cloud.

Which is why I find it so annoying when I see these price comparison articles that don't really use the SpotFleet API correctly.

Agreed. The value of AWS is not in the cost of the instances it is the services and customer service which go along with it. Its all of the easy manages services that add up. Its the combination of RDS, SQS, and all of the other features that add up. Sure each EC2 instance may be more expensive, but there are huge savings in time (which is more valuable) elsewhere.

What services are missing from GCP?

Lots. AWS is very featuresome. In terms of publicly consumable features, they have a 10-year head start.

I'm asking for specific examples.

Does anyone know how much of Google Cloud works inside of China? So, would my webapp be accessible from the PRC? And what about services that come from a Google API such as voice-to-text or translation, managed database and so on?

ah, I almost forgot! My gce instances weren't accessible from Hainan. They aren't accessible from Crymea, also.

I've used GCE for production servers in the past and was really pleased with almost every aspect. I have really little "serious" experience with AWS, and I'd say it's much more.. um, arcane?

I'm genuinly curious, what kind of project/situation would be to actually prefer AWS to GCE (aside from "we're already on AWS" or "I know AWS and don't know GCE")?

Is there a spot instance-like mechanism with google cloud? I regularly request high memory spot instances (r3.8xlarge) on AWS for 1-2 days processes (genomic analysis). With spot-pricing it can be pretty cheap.

A very clever way to run genomic analysis on GCP is via WDL ( https://software.broadinstitute.org/wdl/ ) workflows executed on the Cromwell ( https://github.com/broadinstitute/cromwell ) service. Using WDL you define tasks, and compose those into workflows, then Cromwell executes those tasks on GCP using the pipelines api ( https://cloud.google.com/genomics/reference/rest/v1alpha2/pi... ). So you don't need to set up any instances at all, each stage of your pipeline, a task, is executed in its own docker instance, with files from google cloud storage being materialized on that instance for you. When that task is done, outputs are sent to cloud storage buckets, and everything is torn down.

Picard can pull reads directly out of the google genomics API, and more of the Broad tools are adopting the GA4GH api, so this means an increasing amount of core tools anyone doing genomics will want to use, will work directly with google genomics.

Yes. Google calls them "preemptible instances". Unlike the AWS spot instances, they're fixed price at a discount.

Preemptible Instances only can be used for a max of 24 hours, so would not fit the above usage pattern...though I don't know if AWS spot instances are really meant to be used for that long either. Of course both allow you to use the regular non-preemptible on-demand pricing.

https://cloud.google.com/compute/docs/instances/preemptible

Cloud Genomics is intended for use cases like yours: https://cloud.google.com/genomics/

https://cloud.google.com/preemptible-vms/

Good article though I would disagree about AWS Reserved Instances. Though they are way more complicated, they can give 40%+ discount and in many use cases you can run quite a lot of them.

Isn't it true that those require a 1-yr or 3-yr commitment? You can't move a reserved instance to another region. IIRC, its tied both to a region and AZ.

OTOH, spot instances can be cheaper but then you should be prepared for those to vanish anytime.

As of September 2016, you can buy a convertible reservation that allows you to move a reserved instance to a different AZ within a region. Your greater point still stands but I just wanted to correct one piece of it.

Source: https://aws.amazon.com/blogs/aws/ec2-reserved-instance-updat...

Thanks for the update. The change appears to be relatively recent, and I'm not surprised being unaware :)

Once GC has a RDS like service for postgres, I will switch right away.

In response to Reserve pricing is bs..

We are moving to aws before the end of the year and it will be a 50% price cut from the dedicated providers we have been using. Our dedicated servers are just about out of space and their bs cloud environment is about 350$ for what you get for 70-80$ on linode. When we need 10-50 processing servers quickly cloned up we go to linode do our work and try to sling the data back ...it's all too annoying now. We're moving to aws, reserving some instances for our client front ends and and booting up what we have to when we need it for mapreduce jobs. For us reserve pricing isn't so bs.

It gets a lot more challenging once you outgrow your initial reservations. Or maybe you find that you reserved the wrong sizes. As you add more reservations, it will become something that you are dealing with constantly.

It's terrible.

>Or maybe you find that you reserved the wrong sizes

You can resize RIs within the same family (ie 1 m4.xlarge to 2 m4.large)

http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ri-modify...

Yes, you can. But doing this is not "free", in that you are spending time doing this (potentially often) instead of an infinite number of other more productive things.

This alone isn't going to kill your productivity, but it's one of many of the annoyances of managing a large fleet using RIs. It's a cumulative burden, between figuring out reservations per-region, consider your baseline vs spot/on-demand levels, potentially put some other stuff you don't need on the RI market, and audit how many RIs in your inventory are actually in use.

This is an annoyance when you have a small fleet, but it quickly becomes an expensive nightmare when you start talking 100+ VMs. You start having to pay for expensive external tools to manage your expenses because this stuff sucks so much. Or write a ton of your own tooling (which is not free at all). Or just have someone who does this manually all the time (also really sucks).

Does anyone know if either Google or Amazon provide credits for non-profits like Microsoft does [1] for Azure? I haven't been able to find any details but also not sure if it's something they do offline on a case by case basis.

[1] https://www.microsoft.com/en-us/philanthropies/product-donat...

It looks like AWS has a credit program for non-profits[0].

GCP has one for education[1], and a program to use GCP to protect journalists[2], but I'm not seeing anything for non-profits. You may be able to get credits by asking them, however.

[0] https://aws.amazon.com/government-education/nonprofits/

[1] https://cloudplatform.googleblog.com/2016/06/new-Google-Clou...

[2] https://projectshield.withgoogle.com/public/

Thanks! Not sure how I missed that Amazon page. I'll get in touch with them since having access to Postgres via RDS is very appealing.

While most will still use AWS, it is great that there are alternatives and competition. A lot of good points raised here about Amazon as ecosystem.

The author of this article mentioned running a database server on raid 10 local ssds. I would whole heatedly recommend against this. There is no replication on locally ssd ( at all ). If there is a massive hardware failure there are still situations were all the the data on the local ssds can be lost.

( Disclosure: I would for Google Cloud Platform Support and have seen this happen on rare occasions.)

GCE is a great platform with many competitive advantages including networking, flexible configurations and sustained use discounting. However, I'd argue that using different data points and comparison criteria the opposite hypothesis could be made in favor of EC2 - e.g. burst instance/storage, spot and reserve pricing, local SSD and magnetic storage, and others.

I agree, you can go into higher granularity a-la-carte on both fronts.. but I don't think "spot pricing" is that, and neither are local SSD + magnetic storage (take a look at my local SSD post in this thread).

I would use Google Cloud Storage or S3 Storage but their Traffic charges are HUGE.

500 GB Storage (Google Cloud Storage Nearline) Storage Fee= $0.01 x 500 = 5 Dollar Retrieval $0.01 x 500 = 5 Dollar Bandwith Charges $0.12 x 500 = 60 DOLLAR

I mean SIXTY DOLLAR just for downloading my backups, they are crazy. Both companys.

Would use rClone for my backups but OVH OpenStack is kind of buggy and could not get it to work.

The traffic is free if it's coming from an instance in the same region.

This is really intended for business. Definitely not a good fit for a personal backup solution.

Well i would use it as server backup but in this moment i got rclone woth OVH running :D

https://www.ovh.com/us/cloud/storage/object-storage.xml

10 times cheaper than GC oder AWS

I'd suggest Arq + Amazon Drive.

Its Server Backup

How could I apply for startup credits if I am bootstrapping?

"To Apply, contact your VC, Accelerator, or Incubator and ask about GCP for Startups application details." [0]

[0] https://cloud.google.com/developers/startups/

GAE has serious problems: 1. security, everything accessible for everyone 2. billing, it stoped accepting my card an your ago without any notice, hover the same corporate card works for all other services 3. support it's 100% per cent useless. They cannot even fix anything, only claiming they have the same issue for the last 3 years (that's from my real conversation).

That's why I highly not recommend to have nay relationships with Google Cloud. It's unpredictable, you just cannot build your business around it. Now it exists, in 5 minutes they may decided to close the service, change the billing to unbillable, whatever. And the most scary, NONE, just NONE helps you. It will be your issue.

This is good news for GCS but they will need to compete on more than price if they want to gain market share.

Saving 500$ / month is minuscule compared to the time it will cost engineers to migrate from AWS to GCS and get used to the new service.

Look at Terraform. With the right infrastructure tooling, it's not hard to move at all.

Look at Ansible (or Puppet, Chef, Salt, etc.) With the right configuration management tooling, it's much, much easier to migrate services.

The web interfaces for Google Compute Engine/Disks/Networking are much more intuitive. The pricing calculator is built right in. And - they offer free consulting services to help you migrate out of AWS. We are moving out of AWS to Google Compute Engine and it has really gone fairly smoothly and within a couple of weeks was already saving us money.

Is there another source for the 220Mbit/s limit for m4 class instances? I was under the impression that you could get enhanced networking with those, and double the speed when paired with placement groups.

The 220Mbits/s is for the m4.large instance, not all the m4 family. The bigger instances get more.

It is with HVM and enhanced networking. You can "wget http://ipv4.download.thinkbroadband.com/100MB.zip" and see what speed you get.

The 220mbps limit is entirely false.

Between two m4.large in two different AZs.

Test Complete. Summary Results:

[ ID] Interval Transfer Bandwidth Retr

[ 4] 0.00-60.00 sec 3.13 GBytes 449 Mbits/sec 39391 sender

[ 4] 0.00-60.00 sec 3.13 GBytes 449 Mbits/sec receiver

CPU Utilization: local/sender 1.4% (0.0%u/1.5%s), remote/receiver 7.2% (0.8%u/6.4%s)

This is why Werner Vogels sweats so much.

Both AWS and Google Cloud does not support Ipv6. Suprise Surprise

I use both clouds extensively while google cloud is cheaper in most cases. But DynamoDb in AWS is still cheaper than Cloud datastore or big table

sorry, but parts of this article are a bit.... dumb.

for example, "minimum production instance", it's comparing a 2cpu aws instance vs a 1cpu gce. no wonder its "50% cheaper".

I use GCE over AWS, and run aprox 50 vm's. GCE is cheaper, but not nearly as much as the article claims. the savings is for hardware. data egress cost is basically the same.

i recently moved a database off of RDS onto a "bare metal" VPS and my queries are > 10x faster

The author references this post: https://plus.google.com/+RipRowan/posts/eVeouesvaVX

But quite obviously only read the headline.

Personally I just turn things off when they are not being used and that alone saves me 60% off my AWS bill. Obviously, it only works for non-prod instances but that is the bulk of what we use. see www.parkmycloud.com

last time I explored my options, setting environment variables (ENV) were hard to secure in Google Cloud, I had to push a config.json file that sits on the server? Whats the point of a cheaper price if I can't even figure out how to secure the app. There might be cost gains in larger apps but for the small app I am working on, I'm a big believer of using tools that make you productive and iterate over parts that makes sense. You shouldn't choose a technology like Go lang because its cheaper to run than Node.js if you don't know how to Go

I work at AWS and I have questions for the author about this post:

TL;DR -- I urge all readers to take this post with a grain/boulder of salt: The author is anonymous, prone to hyperbole and error, and makes multiple unverified claims. I'm sure he's a nice guy/girl and if we met in real life I'd be happy to converse over a beer -- but I find the language and misinformation in the post overly polemic and disingenuous. If this stuff interests you guys tune into the reinvent livestream next week: https://reinvent.awsevents.com/live-streaming/

All numbers are from EU-WEST-1 (Does anyone know why GCE doesn't have a eu-west-1a? only b,c,d? I'd be curious to know the story there... not trolling -- just curious.)

Graph 2: c4.large has 2 CPUs... n1-standard-1 has 1... comparison on price is strange? Your point below about optimizing for manageability doesn't really make sense to me as manageability would stem from config/deploy/etc. -- not from instance type? Perhaps your language isn't clear. You're anonymous but I'm guessing english is a second language for you (judging from sentences in post) so it's possible we're missing your point there. Could you clarify please?

>"an ancient virtualization technology" source? details? KVM came out in 2007. Xen came out in 2003. I don't consider either of those dates particularly ancient but whatever it's your post, use the language you want.

Graph 3: A c4.4xlarge has 16 CPUs, and 30 gigs of RAM. An n1-highcpu-4 has 4 CPUs and <4 gigs of RAM. Comparing the two on price is disingenuous. Your claims that these are the "production instances" don't make sense to me.

>Network Heavy: So this is a test between two t2.micros in different AZs: https://s3.amazonaws.com/ranman-code/2016-11-20+00.58.23.gif

It shows 1gbps... I ran it literally a minute ago...

So that's 1gbit... Do you mean inbound from public internet or outbound to public internet? I created an n1-highcpu-4 and had it talk to a c4.xlarge at 1gpbs in (both in eu-west) so your claim that you need a c4.4xlarge for 1gbps seems dubious.

Test Complete. Summary Results: [ ID] Interval Transfer Bandwidth Retr [ 4] 0.00-60.00 sec 7.40 GBytes 1.06 Gbits/sec 1297597 sender [ 4] 0.00-60.00 sec 7.40 GBytes 1.06 Gbits/sec receiver

It actually seems like Google's network was the limiting factor there because when I ran a similar test on a 10gpbs instance I couldn't get faster than 1gpbs. Which is fine because that's what's advertised -- just pointing out that the need for a c4.4xlarge is wrong. On a c4.large I got 900mbps.

>C4/m4/r3 have a hard cap at 220 mbits

This one is just false. What am I missing here? Where did you get this info? From eu-west-1 to us-east-1 I get faster than that using public routes and t2.micros. Internally across availability zones I get much faster than that.

Local SSD #s: Use PIOPS volumes, tunable size, lower cost, can attach to any instance type. Also consider the numerous other instance types that offer ephemeral SSDs.

>It is quite flexible. For instance, we could recreate any instance from AWS on Google Cloud. I don't think you mean any instance... but ok sure!

>Amazon does everything wrong, and Google does everything right, A message by an employee from Amazon than Google, not directly relevant but still a good read.

^ I don't think you read the above post when listing it as a reference. The title is ironic. Steve Yegge quit Google. Famously... on stage... He also goes on to say that Google can't do platforms. I'm sure that's changed in the past few years though.

In the end I'm super excited about both AWS and GCP they both have awesome products. I encourage people to continue researching and writing posts around these topics. It's hard to not take some of the criticisms personally when you're passionate and invested in your work (at least it is for me). I've got a google employee DM-ing on twitter calling me "petulant child" among other things. We don't need that. It's not helpful for our companies and it's not helpful for our customers.

I'll encourage everyone to tune in to the reinvent livestream on the 29th, 30th, 1st: https://reinvent.awsevents.com/live-streaming/

Until bandwidth pricing is fixed rather than nickel and dimeing us to death; a lot of us will choose fixed pricing alternatives to AWS, GCP and Rackspace.

Note... I work on GCE. That said, numbers I'm posting here for GCE are from plugging custom instance and volume types into the GCE console on my personal account rather than anything formal (I'm on my personal laptop). While you can nitpick the OP's points, our custom instance types make it possible to do something closer to an apples-to-apples comparison, and by my math below (which may be wrong; the risk of posting at 1am) the 50% number still holds up if you ignore reserved instances, assuming the OP's prices for AWS were correct.

> Does anyone know why GCE doesn't have a eu-west-1a? only b,c,d? I'd be curious to know the story there... not trolling -- just curious.

Just guessing (I've worked on GCE for a bit over three and a half years, but I don't remember the history here) -- I suspect that it was a zone that existed prior to GCE going generally available that was turned down and there was a desire to not re-use the name.

> c4.large has 2 CPUs... n1-standard-1 has 1... comparison on price is strange?

Well, ok, but a custom 2 vCPU instance with 3.75 GB of RAM is $45/month (including a 10 GB PD root volume), so still nearly 50% lower than the c4.large price.

Similarly, if I build a c4.4xlarge equivalent custom instance type (16 vCPUs and 30 GB RAM) it's $358/month, again putting it at just about 50% of the rack rate for a c4.4xlarge.

I'm not sure where the OPs claims were coming from for networking. Certainly you can get 1 Gbps out of relatively small EC2 instance types, although at a virtual NIC level GCE is considerably more generous, providing virtual 2 Gbps/s/vCPU link speeds (so up to 8 Gbps for a 4 vCPU instance versus 750 Mbps for a c4.xlarge). Bandwidth-to-the-edge will be lower. That said, when trying to find an upper bound on bandwidth it's really best to test (regardless of which platform you're evaluating) with multiple streams ideally to multiple peers. Single stream can end up limited by a number of factors including RTT and congestion window scaling bounds as well as getting unlucky in terms of ECMP path selection for multi-path links.

> Local SSD #s: Use PIOPS volumes, tunable size, lower cost, can attach to any instance type.

PIOPS volumes are EBS-SSD, not local, aren't they (looking at https://aws.amazon.com/ebs/pricing/)? Our analog would be PD-SSD. GCE PD (including PD-SSD) IOPS scale with volume size — let's say you needed 20 GB of storage at 10000 IOPS. On EC2 you'd be paying $0.125×20+$0.065×10000 = $652.5. On GCE you'd need to buy a bigger volume, at 30 IOPS/GB for PD-SSD you'd need a ~334 GB volume. However, at $.17/GB you'd get the same 10000 IOPS at $57/month — for this example EC2 is more than 10x as expensive, and if we scale the storage up so you get the same 334 GB on io1 EBS PIOPS SSD it's more than 12x as expensive.

I picked these numbers somewhat arbitrarily; EC2 would've come out looking better for very large volumes with very low performance requirements, but with PD-SSD priced at $.17/GB and io1 volumes priced at $.125/GB + .065×1IOPS = $0.19, io1 volumes are still more expensive even at the lowest IOPS bound (above 1GB they get cheaper, but only below 3 IOPS). General purpose SSD is cheaper than PD-SSD, so I suppose for users who can live without a performance SLA it might be a good choice, although at that point I'd be tempted to benchmark it against our vanilla PD offering.

You can also attach GCE's truly local SSD (as opposed to PD-SSD) to any instance type (with really fantastic IO performance -- up to 680,000 IOPS read perf with multiple volumes), although I'll grant you that the 375 GB granularity is a little... coarse.

> source? details? KVM came out in 2007. Xen came out in 2003.

To be fair, GCE doesn't exactly run on stock KVM (nor, I assume, is EC2 running on stock Xen), and our userland is not qemu. I don't think it's meaningful to try and assess these technologies from the outside[0].

> Steve Yegge quit Google. Famously... on stage...

He didn't, he just switched projects... he's still at Google. He went on to say that Google couldn't do platforms; one of my favorite things about working here is the company's ability to be introspective and recognize when a rant like Steve's has more than a grain of truth to it. Google adapts. I think comparing the GCP APIs to Google's historical API efforts makes it abundantly clear that Google has improved by leaps and bounds at building a cohesive platforms.

[0]: I do lead the team that owns our virtual NIC and a chunk of our host dataplane in addition to prior involvement in broader aspects of our hypervisor architecture, so I'll admit to more than a little bias on this front :)

Is there a roadmap for GCP - wondering if they are coming to Australia?

2017:

https://cloudplatform.googleblog.com/2016/09/Google-Cloud-Pl...

I find AWS annoying in that the east coast is always cheaper and the price reductions linked mention more reductions on that side.

Who of the cloud companies is treating California as a first class citizen?

Oregon region is on par with east coast.

My problem with Google Cloud is I find their pricing too opaque. Will I find myself with a huge bill overnight is my worry. Is there some way to get started with a cap on my monthly bill?

Bandwidth pricing, bandwidth pricing, bandwidth pricing for all the big cloud providers that charge by the bit rather than a fixed price; all of them are too expensive.

The only feature that any company can copy from a competitor without any effort is price. It is well known in product circles that you can't compete with price only, you need to put in other features that make your product attractive compare to the competitors. I am still waiting for something extraordinary from GCP that would make me consider them over AWS or Azure. Price - for me - is simply not enough.

The other 50% is subsidized by the NSA in exchange for direct access to your servers.

This PR propaganda is making the rounds. Most of the real use-cases I've seen point to self-hosting on AWS is cheaper than using GCE and friends.

Pick whichever service has the cheapest TCO, not what some paid blogger says.

Azure is even cheaper

Concerning the Virtual Machine pricings on the discussed platforms, I'd like to mention that – on very small machines – running Windows Server is significantly more expensive on Google Loud than on AWS or Azure, because Google charges you the license fee directly whereas Azure or AWS seem to have a mixed calculation.

Having the smallest Windows VM on AWS is like 6-7$/month, whereas on GCE you're at at least 14.60$ higher due to the license.

Correcting my comment, Google has apparently undercut Microsoft recently.

well, both still loose to budget app hostings if you count only cost

I wonder how do they compare to digitalocean?

That's not a good comparison to make.

what makes me choose AWS, is i can survive a year using it without paying too much money, while in GC you got 300$ to spend in 3 months, which would get you more powerful machine for this period but you should start paying after it so if you really don't have money to start with, you have to choose AWS

I don't understand. Why not just use a normal amount for the first 60 days.

doesn't matter, it still just 60 days

Just 60 days for what? You get full functionality before and after. And you pay less than AWS before and after.

I test and use all 3 main public cloud providers. I keep Azure out in this conversation for now:

A quick comparison of GCP & AWS :

The services & flexibility of options offered by GCP is not even in par with AWS. Yes I may be able to spin up a couple of VMs but when it comes to enterprise GCP can't be an option. It is like buying the new macbook pro and can't add more than 16gb memory! A designer for sure needs more ram regardless even if Macbook pro is offered for free.

- There are about 50+ 3rd party offerings in Google Cloudlauncher compare to thousands in AWS Marketplace! - Choice of server/OS! maximum 8vCPU-7.2Mem in GCP compare to 192 vCPU and 2TB memory in AWS -NoSQL Big table ( max 30 nodes) - compare to ( virtually unlimited DynamoDB) -SQL (only MySQL in GCP ) compare to ( Almost all with exception of DB2 in AWS). - and so forth - Just having Lambda, F5 support is damn good reason to go for AWS regardless of how much it costs. Those are the basic needs of SMB that google can't even offer. - 60 day trial and 300 dollar limit! compare to 1 year

The whole point of moving to the cloud is to have no limits and be able to have access to resources whenever you want. Google offering is very limited for now. This is just a lame marketing move by google and I hope they first fix their offerings and then compare. I can go to godaddy and say I'm cheaper maybe! Just check the list of services offered by both and see which one should be considered for a serious customer.

We are all professional and judge products by testing them, open an account in both and judge by yourself. When I started using GCP it sounded refined for basic services but when it comes to Bigdata, monitoring I did not find the console very integrated nor usable.

I work for Google Cloud. I think a blanket statement claiming the GCP is 50% cheaper that AWS is a bad for all the reasons you stated, there is a lot more than just VM pricing you have to take into account! We are definetly working on a lot of the things you brought up, I hope this time in 2017 your experience will be very different.

Want to clarify a few things though:

> - Choice of server/OS! maximum 8vCPU-7.2Mem in GCP compare to 192 vCPU and 2TB memory in AWS

The biggest instance on GCP currently is 32vCPU-120GB. Also, with custom machine types you can tune your machine and pick exactly how many cores / memory you need; you are not stuck with predefined instance types.

Curious what OSs AWS supports that GCP does not?

AWS definitely has a better vertical scaling story though.

> - NoSQL Big table ( max 30 nodes) - compare to ( virtually unlimited DynamoDB)

BigTable and DynamoDB are not really comparable. BigTable is much lower level. The apples to apples comparison is DynamoDB and Datastore, and Datastore is also unlimited.

> - 60 day trial and 300 dollar limit! compare to 1 year

I'd also like to see a longer trial for GCP, but the AWS trial gives you 1 year access to basic usage. With GCP, you can spend the $300 any way you like. Personally I'm not sure which model is better or worse, though I lean towards the AWS model.

Reply on news.ycombinator.com