Comment by airstrike | Hacker News Reader

Comment by airstrike | Hacker News Reader

> Neither company is satisfied with that, of course. Google so much so that it moved on from Diane Greene at the end of last year, bringing in Oracle veteran Thomas Kurian to lead the division out of the doldrums.

> (...)

> Bloomberg reports that he announced a plan to increase the number of salespeople and train them to understand specific verticals, ripping a page straight from the playbook of his former employer, Oracle.

I'm not an Enterprise IT expert, but looking from the outside, it doesn't seem like AWS and Azure got to where they are today by following that playbook...

My personal take on it is that AWS was first-to-market with a 21st Century-ready cloud offering (i.e. you could pretty much hear developers shouting "Yes, THIS is what I'm talking about. Finally!") and Azure got to where it is by responding quickly and aggressively, with an additional pull for certain customers due to the synergy with MS / Windows, as well as being the next best choice for those directly competing with or not willing to sign up to Amazon.

If that's true, then unless Google / IBM can really differentiate themselves, they will always lag far behind. I don't know what other plans they have in store, but opening up Watson to other platforms or hiring an Oracle sales guy are not quite what I'd call differentiators.

Azure was pretty late but their main fairly simple task was to cloud-enable the massive .net/MS-SQL army. Initial attempts weren't great (Azure tables, branding changes) but they iterated fast and knocked it out the park.

Google is replicating Amazon now (after failing with app engine), but without the mindshare. Amazing because they are the cloud story - many cheap computers rather than huge ones. App Engine hurt them so much they have a hell of a job to recover and now have to compete using someone else's playbook.

Right now, nobody gets fired for using Amazon.

> Azure was pretty late but their main fairly simple task was to cloud-enable the massive .net/MS-SQL army.

I'd be interested to know about the number of:

* Linux based deployments on Azure * Windows/.Net based deployments on other providers

Basically I'm assuming that all the .NET crowd is hosted at Azure and the rest are being hosted on Amazon/Google/RedHat/Heroku.

If that assumption is true (even with 4-5% of room for error) the ~20% would be .NET's share on the backend.

Am I missing something obvious?

Azure is bought in the boardroom, not by the engineers and ops staff that will use it on a day to day basis.

When you have a $xx million Azure minimum commit that your C-level has already written a check for, you yell at your account team until they add Linux support and try to make due with what you have.

I've heard internal numbers that Linux VMs outnumber Windows 2-to-1, but that they see long term value in doing a "lift and shift" of existing apps, then down the road sell them consulting services to migrate to .net.

Makes me wonder if instead we might see an MS branded Linux distro (think redhat) or C# native on Linux.

They already have both. Azure sphere and .Net core.

> I've heard internal numbers that Linux VMs outnumber Windows 2-to-1, but that they see long term value in doing a "lift and shift" of existing apps, then down the road sell them consulting services to migrate to .net.

Presumably they're making a profit on selling the linux VM's. I think MS have more or less accepted that windows has been outcompeted in the server OS market over the long term. It's hard to compete wih free...

Microsoft does not need to shift a single Azure Linux customer to the MS stack to make money on Azure. This was a major mindset breakthrough at MS that Nadella is not responsible for, but has leaned into. MS is not a Windows company anymore, they are back to being a software company. And Linux is fine software.

AWS delivers most of Amazon’s profit despite Amazon having no proprietary software story to speak of. Well, MS can do that too.

EDIT to add that MS does not even sell consulting services, their partners do. I’m sure the “lift and shift” story makes sense to long-standing MS shops, and MS may have even included that in their Azure pitch to them. But ultimately one of the best things about Azure from MS’s perspective is that they don’t need consulting partners to make money from it. They can reach directly into the product and ops teams at major enterprises and sell them raw capability with a spend that auto-scales.

Actually Azure has a huge hook at the engineering level in that Visual Studio Enterprise subscriptions come with $150 a month in Azure credit. It's very hard to avoid using when you just need to slap a quick demo online and need a server or DB and your in-house IT is away, it's all right there in the IDE... and then you're in hook, line, and sinker. So while there are C-suite level mega-sales, Azure has plenty of grassroots appeal.

That's actually a great point, Visual Studio becomes a distribution point for Azure client software and SDK's, meaning you potentially don't have to do any extra configuration the way AWS and GCP require.

All the .NET projects I have done on the cloud have been deployed on Amazon EC2 instances.

Microsoft said in September 2018 "about half Azure VMs are Linux." [1]

Of course, that leaves it open what they mean with "VM" here. Are they including all the VMs running in Azure, including those powering Microsoft PaaS offerings (like App Service and Azure SQL) or just customer deployed VMs.

[1] https://www.zdnet.com/article/linux-now-dominates-azure/

Right. My only complaint on AWS is how confusing it is. Gcloud is much simpler, IMO.

I've looked into a few AWS service over the last week and coming from Azure I am very confused. It's not at all clear how AWS service are grouped together or why certain resources where created at all.

The resource group concept in Azure is, in my opinion, much easier to use and understand that VPCs in AWS.

Azure is horribly complex too. There is an involuntary comical course on pluralsight [1] trying to explain all the different types of services and the small subtleties between services that do about the same thing, but not exactly the same. It’s hard not to leave even more confused.

[1] https://app.pluralsight.com/library/courses/microsoft-azure-...

AWS is easier to set up than to plan your own data center. The difference is that people who opt for AWS think they no longer need ops, and don't even have knowledge of networking. You still should have at least one network engineer to plan VPC and subnet layout.

As for GCP, I think it seems easier, because by default it uses public IPs for each instance. That doesn't necessarily mean that one you set up an instance and it runs, you did it correctly.

One nice thing though is that Google uses SDN for their internal networking and that made some things easier. Particularly seamless communication with other regions.

Unless your business is in direct competition with other parts of Amazon's service.

It's not that businesses actually fear that Amazon will look at the data they have in AWS, but AWS props up other Amazon businesses, and who wants to help prop up their competitor?

Seeing that AWS’s largest customer is Netflix and that AWS has Prime Video, that’s not necessarily a showstopper. The tech industry is all about competitors who partner.

This isn't true in Retail. As far as I know, most of the top national retail brands in the US all use Azure or GCP as a policy over AWS.

> App Engine hurt them so much they have a hell of a job to recover and now have to compete using someone else's playbook.

why did this hurt them? I thought appengine was great when i used it (granted, i never paid them for it - the free tier was way too generous).

It delayed them making the product customers wanted?

It gave a lot of developers for a long time an impression that Google only really cares about (maybe) Java and Python and other languages are SOL?

It gave a lot of developers for a long time an impression that Google both very much wanted proprietary lock-in (App Engine APIs were extremely distinct to other libraries commonly used for those tasks in those languages), but didn't have the dedication to support a lot of backward/forward compatibility and were happy to change those proprietary APIs seemingly on a whim?

Three things. First, GAE has drastically, massively improved as a product since it was first introduced in 2008(!). But a lot of people don't know that, and the original reputation remains.

Second, App Engine was the original Google Cloud product, it came out years before Compute Engine and never took off like they expected but for awhile it was all they had to offer.

Thirdly, in 2011 Google changed the billing model. Datastore requests were no longer free and CPU billing was switched from CPU time to wall time [1]. The old billing model did not account for how some developers were actually using GAE and it was not sustainable, so Google had to make a correction and some apps saw a massive cost increase. This further tarnished the GAE reputation.

[1] https://github.com/stickfigure/blog/wiki/The-Unofficial-Goog...

AWS employs enormous number of "sales engineer". They are so aggressive that there is a joke that anyone with some name recognition in Korean dev community is going to AWS. I think Google trying Oracle approach is the right one. Their product is not bad, their sales is.

Modern cloud devrel seems very similar to big pharmas marketing to doctors.

excellent analogy

Google needs sales-people on the ground, plain and simple. I live in a reasonably big European market when it comes to IT services and the Google cloud people are absent, and for that reason everybody goes either with Azure (because MS has got the most people trying to make the sale) or AWS (because it's still seen as technically ahead of the field).

Now, as an aside, rumor has it that even MS's Azure numbers are a little inflated, a big part of their cloud sales is just Office 365, but I think MS does some marketing and accounting gimmicks on the side by "attaching" Azure to these Office 365 sales.

My anecdata is that Microsoft used salespeople to talk to executives and give them a deal on Azure so good they couldn't pass it up.

Migrate your Exchange serve(s) to google suite or Azure, which one do you think is an easier sell?

Repeat for Random windows server to Azure and again for Office Live.

Ohh, you can ru Linux too and do most things people are doing with AWS.

It’s a pretty easy sell to me (and worked pretty well on the F100 company I work for).

Microsoft is pretty clear that they want people to move workloads to PaaS and SaaS - with IaaS (VMs) really only for "legacy" applications.

The migration from on-premise WSAD and Exchange to Azure AD and Office 365 is pretty painless - and once you've done that why wouldn't you use the rest of Microsoft's cloud based platform. That's the big advantage Microsoft has for most enterprise customers.

This is occurring at my company. That said I'm not too sad as Azure seems like a competent provider.

I'm not terribly happy at the VM startup latency, unreliability of actions (you have to double check if it really happened), and how things don't clean up nicely afterwards. Things may have changed, it's been a few months.

Yes, this (at the top level) was how they signed up fortune 1.

This is accurate. If you have a substantial Windows Server / SQL etc. spend via an Enterprise Agreement then MSFT will essentially give away their cloud services to you. I’ve seen crazy 90+ percent discounts on Azure and Office 365.

Still doesn’t change the fact that Azure isn’t a great platform to use, though.

I have an affinity for AWS because I worked there and liked the pace. But, agree with your last statement. At least 3 collegues (came from Azure to AWS) told me very highly about Azure when it comes to console and workflow.

I think aws also did a great job keeping competition "under fire". They release so many new services, so quickly (and in amazingly alphaish-at-best quality), that they are always perceived to offer the most complete cloud infrastructure.

I was playing with gcloud, particularly with deployment manager and have feeling that there are still many bugs in gcloud. DM seems like it has a good design and could be more powerful (especially ability to use templating and injecting Python code), but still feels buggy, error messages are very cryptic, and not feature full around updating (there is some rollback and preview functionality, but it is essentially useless).

Nowhere close to what Cloud Formation is, even though from design it is less powerful.

I was a bit skeptical why terraform is popular, doesn't feels like it brings anything new to CF, but could make a lot of sense to use instead of DM.

Even being a differentiator doesn't seem to help. Joyent's Triton infrastructure is extremely impressive and container-centric, yet hasn't made a major dent into AWS.

Joyent's failure was chasing the latest agile, devops, micro kernel, etc. hotness that was taking off in the valley.

They completely ignored the fact that the big companies with huge budgets are full of people who don't give a shit about SmartOS resource grids and just want easy and similar enough to what they already know/run.

People tried to label Amazon Lightsail as a "DigitalOcean killer", it was really a way to lower the bar of intimidation for the IT guys at paper mills and car seat factories.

Reply on news.ycombinator.com