Google Cloud CEO says that he’s borrowing from the Oracle playbook
businessinsider.de
businessinsider.de
As a Google Cloud customer, I really hope that he doesn't take the "bad bits" – raising prices, unnecessary lock-in, slow moving technology, litigation. As an engineer in a team where engineers lead decisions like which cloud provider to use, rather than some exec on the golf course, I hope they keep their appeal to those customers who don't want to play the enterprise sales game.
Really? This is interesting. I am not working in a large org, but all of the issues we've had with their user management have essentially come down to it being designed for companies much larger than we are, with much more complex security requirements.
Can you elaborate on what doesn't scale? What sort of requirements are they not meeting? Is it _targeting_ large orgs and getting it wrong, or is it targeting too-small orgs in your opinion?
the default service account is a project editor, which is insanely over-privileged, yet everything tries to use it, and all the demos. Appengine only supports one service account for all apps, even if you have multiple apps, also Editor. similar things in gke.
Their pre-defined roles are a mess, and not consistent across services. Theres a role called 'dataflow admin' defined as the 'minimal permissions to run a job' which isn't what id call admin.
So if you try to use custom roles instead, cause the predefined roles are crazy, they will make breaking changes to permissions(literally did this last week with no notice).
Support access doesnt use IAM and only supports users, not groups, and no api - you have to go to a web form and add each user individually. good luck if you have more than a couple users, and have fun trying to keep that in sync.
The fact that its tied to gsuite is a problem as well, a gsuite admin is also a gcp admin, which is generally owned by different orgs(IT vs dev).
it seems like every month we have a new breaking change, or price change(maps is biting us constantly for some reason)
ultimately what it comes down to is weve been burned enough that we dont trust google
We have hit a few of these. I hadn't even thought about the GSuite/GCP crossover – that's kind of good for us, but yeah in a larger company I guess they would be totally different orgs.
Some of it is cowing to typical enterprise asks, like redlining objectionable contract terms, willingness to deploy on-site consultants, supporting versions for longer time horizons, being more flexible with egress charges since enterprise places can't migrate all at once, offering experts up to your finance org to explain how to make cloud compute a capital (vs expense) line item, etc. Things MSFT is very good at, AMZN at least tries, and GOOG historically has ignored.
That's a small sales force. You can build solid products, but you've got to have a sales force to go with it. Amazon had advantages as the first big mover in the cloud market. If you're coming in later, you've got to be able to sell.
Google is DigiKey, Amazon is Adafruit, and Microsoft is Best Buy.
Based on:
https://www.skyhighnetworks.com/cloud-security-blog/microsof...
AWS holds 41% of the market. Azure 29.4%. GCP 3%. So barely more than a tenth of the second place cloud's size.
For a company with Google's brand recognition, and a cloud platform that has been going for that long, that's not great market penetration.
Microsoft adds Office 365 etc to Cloud numbers, G Suite is a separate line item.
Also, Microsoft has other products that neither Amazon nor Google even provide an equivalent for (such as Dynamics NAV), also added to Cloud numbers.
Here's another analyst report from Feb 2019 that estimates shares based on dollars of revenue at 32.2%, 16.5% and 9.5% respectively.
https://www.canalys.com/newsroom/cloud-market-share-q4-2018-...
I'm wondering if their practice of dropping products isn't coming back to bite them.
Any report comparing the 3 purely based on revenues should be taken as a joke.
The basic instance n1-standard-1 vs c5.large is 80% more expensive on AWS. A shame AWS doesn't provide single core instances.
Google gives 30% discount automatically on instances running for a full month. AWS has no such thing. That's enough to put them in different league.
Anything AWS involving SSD or provisioned IOPS is abusively priced whereas Google simply has better performance on network storage and can attach any amount of local SSD to any instance.
It's hard enough finding Postgres SAs these days as it is!
That only seems to be true for places unwilling to pay reasonable amounts of money.
And by "reasonable" I'm not meaning Oracle Consultant $US3k+/day either. ;)
Generally I try and steer them to either the PostgreSQL jobs mailing list (https://www.postgresql.org/list/pgsql-jobs/), or if they have specific needs then I'll steer them to appropriate people (if known).
Of the people that contact me... some are Weird (not in any positive way). When I ask them how much they're looking to pay (even roughly), they either avoid answering that point, go silent completely, or just do other strange stuff.
Those ones don't get put in contact with anyone. Not sure what they think they're doing, but wasting people's time isn't on.
https://insights.stackoverflow.com/survey/2019#technology-mo...
Why would I recommend google to anyone when there is no stability in their platform?
What I was getting at, though, is that if you're a big user of GCloud IaaS products and some of their more basic infrastructual SaaS offerings (Cloud SQL, Bigtable, etc. etc.) you're unlikely to run into the kind of issue you mentioned, and everything from the pricing to the way customer relationships are handled reflects that.
Google Maps is basically an application as a service, which, while technically all part of "cloud", is not really part of the core cloud offering in the same way. On the Cloud Console, Google Maps is listed in the menu at the very bottom, as the only entry under "Other Google Solutions".
All the cloud providers have a mix of offerings that range from very standard (IaaS, or managed open source systems like Cloud SQL and AWS RDS) to very proprietary (things like BigQuery, DynamoDB.) If you decide to use one of the very proprietary ones, you have to recognize that you're creating a very different kind of dependency and risk - migrating to another provider or hosting onprem becomes a rewrite, instead of a migration.
That's a risk you should ideally be aware of before you choose such a solution, and it's not unique to Google. That same level of risk doesn't exist for the other kinds of products I've mentioned, so it doesn't make sense to treat that risk as though it applies to all Google products.
The business sets the budget and the requirements, but whose recommendations do you think they listen to when delivering that? Often it is in-house tech' people.
There's definitely a lag between technical people getting interested in something and that thing getting big in enterprise, but there's still a correlation. Just as there's a inverse when everyone in tech hates a thing and a few years later businesses stop adopting it (see Oracle's slow decline, particularly for new customers).
Have you ever worked at a large enterprise? That's almost never how it goes.
An executive has a meal with a sales person, who says their technology will solve all of the executive's problems, then the executive signs up for a paid trial, and then tells the people that report to them that they are going to implement the thing.
Some better enterprises will cut it off at that point after the technologists come back and say the trial didn't work out, but many enterprises will simply plow forward, either falling to the sunk cost fallacy, or because the executive's friend at another company told them it worked well at the other company.
Unfortunately for them they also burned way too many bridges with technologists as well.
I don't get the allure of AWS, but Azure definitely knows how to hook into Enterprise infrastructure. I don't know what GCP "doesn't have" for the Enterprise.
They lack absolutely everything that makes a successful enterprise company.
Customer service. All the way to the CEO/CTO level, where in event a named customer has a problem and fuming at something, the execs show up at customer's site in person to butter the relationship. Never underestimate how important it is to have a high up person with authority to reassign priorities at a vendor do the executive customer service in order to keep and grow enterprise side healthy.
As if Google is teetering on the brink? Or are you suggesting they'll just pull the plug on GCS? They very well might. Which, in fact, is probably half the reason companies won't buy what they're selling.
This is something where Microsoft has done well because it’s always done this and wasn’t too late into cloud. It’s an area Oracle have done well pre-cloud and are late to cloud but hoping like hell to bank on their strong enterprise sales capabilities (doesn’t seem to be working for them though). And it seems to be an area that Amazon have done well due to early mover, being dominant, and with that focusing on building enterprise sales.
Google... I don’t know what they’re doing, they’ve never been an enterprise software company. So he may be into something.
Any startup choosing google cloud is taking an unnecessary risk.
Google Cloud sucks at enterprise sales. By all objective measures, Oracle absolutely owns this part. It's not as black and white as they need to become Oracle in order to learn some things from them.
As someone who consults with a lot of enterprises and helps them with their cloud architecture, I can see first hand that the market share of Google Cloud there is about 0. It's only internal, toy projects and nothing critical, because they already have AWS and Azure and that's that.
Even with all their technological excellence, it doesn't matter. That strategy failed, and they're losing.
I'm willing to wait and see how a sales-centric strategy works out.
https://hackernoon.com/why-and-how-we-left-app-engine-after-...
Enterprise adoption != developer adoption.
That's why Google desperately needs a strong sales team.
As a heavy AWS user, this part actually excites me. Avoiding vendor-lock-in is something I value highly. I would also be a lot more willing to try GCP if I didn't have to invest effort into building a GCP-specific stack. This would also be a huge win for all cloud customers by making the marketplace more competitive.
The sales guy was/is pushy but tbh the tech is very prescriptive on how things are done, it's the Google way or it doesn't work. That is not how enterprise works, you need to adapt to each customer.
We have lots of AWS and plan on using it more, we might be doing it sub optimally but it works.
Either do it right, or don't use us sounds very good.
And if their "doing it right" really is right they'll end up with better outcomes for customers who stick with them, which will drive them ending up with more customers.
> Weeks into his new job as head of Google’s cloud business, Thomas Kurian identified a chief complaint from big business customers: They often didn’t have account managers to call.
Oracle is famous for getting customers to pay, but they were also very good at helping customers with services, advice and customizations. Expensive Consultants worked closely with dev teams. Yes they were expensive, but clients knew they could rely on someone from Oracle fix bugs or give guidance.
https://www.wsj.com/articles/whats-been-lacking-at-googles-c...
Not sure how true that is. Plenty of horror stories around of Oracle bugs not being fixed.
The above does not necessarily apply to bugs and feature requests, just operational support.
Definitely have heard cases of that happening, with large enough accounts.
But also plenty of stories of problems being ignored, same as Google Cloud has a reputation for. :(
(for those wondering: we are being a tad facetious as everyone who has been in the industry a while is familiar with stories of Oracle being litigious, one case specifically being a large and long-running one against Google over Java and Android: https://en.wikipedia.org/wiki/Oracle_America,_Inc._v._Google...)
They are investing greatly in Kubernetes but it seems still a long way to go before the lock-in is achieved, so they can use more Oracle tricks.