67 comments

[ 3.1 ms ] story [ 107 ms ] thread
At the mean time, there are 7 GCP services currently under outage. --A frustrated engineer
I feel better, because I have only one. wonder why gcloud is so bad? not even as good as digital ocean from a user perspective.
Digital ocean has ~300 employees (circa 2017 according to wikipedia). Google has far more than that working on Cloud. Google Cloud is also largely layered on top of their existing internal infrastructure. Google cloud also has more services, more regions, and more zones. My opinion is that there is just way more that can go wrong.
I move people from AWS to GCP for cost, simplicity, and developer experience.

What do you find so bad about GCP?

(comment deleted)
To be fair, these were not outages, rather degraded services where a feature or two was not working. None of my applications had any errors, nor were my users impacted.

You can find the details here: https://status.cloud.google.com/

You also have to account for the incidents that customers face that don't show up in the status dashboard until a certain influx of support tickets raise the flag.

For example, right now we're facing an issue where one of our GCP HTTPS load balancers is suddenly failing to connect to any backend service for reasons unknown. Not sure if there was any correlation to the eight reported service disruptions on the status dashboard.

Yeah, there are still per customer issues that don't make the status page. GCP does a good job with their RCAs, in-depth with future mitigation strategies. They are also really good about keeping the status updated.

I had an experience on both GCP and AWS of disks failing. For AWS we had to manually reset the machine, for GCP, I only know because I looked at the audit logs. GCP auto-migrates on hardware issues.

Sounds like something GCP support would love to help you troubleshoot. The disruptions I'm aware of have no correlation to load balancer traffic paths.

What's the actual reason listed for the 502s in Stackdriver? That's a great first step.

All Stackdriver logs for the load balancer have the following for each request as the status details: failed_to_connect_to_backend

We've tried the following with no luck: - Restarted GCE instances in backend service - Recreated the backend service - Recreated the HTTPS load balancer - Switched the balancing mode of the backend service to rate instead of utilization since monitoring was stating the backend service CPU utilization was 500% (another red flag)

We are, however, able to directly connect to the GCE instances within the backend service via external IP.

Have not heard much back from Google support.

(comment deleted)
Google includes GSuite revenue, which not long ago, was half of all "Google Cloud Revenue". I also assume a large portion of Azure revenue is O365. So Amazon's lead is a bit more impressive than it might seem at first.
It is, but considering the estimated $1T market and all players having some exponential growth-looking curves (IIRC) the difference is pretty insignificant.

I'll hold on to my stock in all these companies for a bit longer. I don't think there'll be any losers in this race.

But startups use Oracle according to Oracle...
Sounds like every deck pitch from a non profitable startup trying to get another round of funding...

“The market we are in is projected to be worth X. If we only get Y percent our revenues will be Z”...

It would, except the tapped market is 30+11+8 billion dollars today, and honestly do you doubt that the market is anything less than hundreds of billions? There are a lot of companies out there who should focus on their business model and not be in the DC-running-business.

Also, Google, Amazon and Microsoft don't need your funding, thank you. They're very profitable already.

I’m not talking about funding, I’m more referring to just because the market is big, doesn’t mean that GCP can take advantage of it. Amazon has first mover advantage and Microsoft is entrenched in the Enterprise. Just going by the old adage of “No one ever got fired for buying IBM”, AWS is the safe reputations choice because “everyone” uses AWS and Azure is safe because enterprise companies have decades of trusting Microsoft.
What I'm saying is that the exponential growth specific numbers matter. The current division doesn't. And also if GCP "only" ends up with 10% then that's a lot.

Amazon could have the best exponent. Could be. Doesn't make the other two losers.

> It is, but considering the estimated $1T market and all players having some exponential growth-looking curves (IIRC) the difference is pretty insignificant.

Last article I saw (which, weirdly, spun this as Amazon losing dominance to Microsoft) had AWS with twice MS’s cloud year-over-year growth rate (that is, twice the base in the exponential formula) as well as a bigger current slice of the pie.

Exponential growth doesn't make differences insignificant if the rates aren't equal (though I guess the starting difference is less important than the rate difference over the long term.)

I agree entirely. My point, in fact, is that the exponents base is all that matters, and current market share doesn't. All three are able to take the whole cake if they sign the customers, and so it's up to the exponent base to decide how the $1T is split up.

That is, all three have the logistics, infrastructure, and partners in place to scale up if they need to. And they don't have to worry much if DigitalOcean has higher exponent base since DO wouldn't know what to do if they were offered a $10B deal. The big three would.

According to this[1] the azure number ($11B) is an estimate as it is all apparently reported under the umbrella of the larger "commercial cloud" number -- so the CC number includes o365 but the azure estimate apparently does not.

30+ vs 11+ though, is still a very impressive lead.

[1]: https://www.zdnet.com/article/top-cloud-providers-2019-aws-m...

O365 consumes Azure for many services. I always wondered how they account for that revenue.
The real question is why Amazon doesn't have email/calendar/chat/video conferencing; "AWS for Business" or something. I am sure they could get a good deal on the hosting costs!
I'm curious why Chime, which arrived with great fanfare, never really took off. I've been on conf calls with lots of different corporate customers, and as such have had to use lots of different clients (Zoom, Skype for Business, Google Meet, Bluejeans, etc.) but never once have I had a customer use Chime.
Amazon uses Chime internally. It gets the job done, but the UX is garbage compared to Slack.
I suspect they could fix it, but Jeff, Andy, and Barr all think it's good enough and noone cares about UI.

Look at Amazon and AWS's web page...

I love Chime. It reminds me of Campfire/Propane with the ability to do calls, video, and screen sharing. I used Slack for a while before joining Amazon and I don’t miss it or the culture around its use at all.
We had one vendor use it. The UX is not good and call quality is just okay. I think it isn’t adopted more often because organizations that want to use something new choose way better options and everyone else is afraid to choose anything unproven/weird or already has established accounts with something else. And it really doesn’t help that Chime sort of looks like an AWS service instead of a consumer product.
It’s a crowded market. Skype for B is free. People tend to graduate into Zoom these days.
I think AWS has been working on an offering like this for a long time. They sue a former exec for joining Smartsheet because it competed with a similar AWS initiative. https://www.geekwire.com/2017/amazon-resolves-contentious-no...

But as a former Amazon employee, the only thing that I can say is that if it's anything like Chime in terms of quality, then Google and Microsoft have nothing to worry about.

Have you used Skype for Business?
(comment deleted)
I’ve wondered the same thing about Mozilla. I feel like should be building more value added web services that go beyond just little add-one to the browser itself. First, they would help sustain the org. Second, they would help encourage use of the web as a platform.
Amazon does have WorkMail [1]. Per the blurb, WorkMail is "secure, managed business email and calendar service with support for existing desktop and mobile email client applications"

https://aws.amazon.com/workmail/

Edit: Looks like the OP of the post has already mentioned this below. My bad!

Microsoft includes Office 365. In a way Google has a more "legit" claim to include GSuite, since their offering is entirely cloud based, whereas with Microsoft only the parts of Office that nobody actually uses are cloud based.
These are cloud products with subscription revenues that are sticky with predictable churn.

Never understood why people devalue them. MSFTs turn around and valuation are based on the O365 pivot. Very impressive.

I wasn't trying to devalue them. Just noting it's not exactly apples to apples if you're interested in a market share comparison.
GCP Cloud Composer (Airflow) has been a nightmare. I cannot recommend it. BQ is great.
GKE is amazing too, none of the other major clouds even come close.
I found Airflow for a small to medium project to be a nightmare on its own; I'm sure its a powerful framework, but plain Python ended up meeting our needs more than adequately. I was self-hosting it while kicking the tires and found it to be a lot to admin. I always thought I would look at Cloud Composer if I wanted to revisit Airflow. Were your issues with Airflow or Cloud Composer? Agreed BQ IS great.
My issues have been with cloud-composer. I like Airflow.
Would love to know the pain points so we can make it better - Composer, Airflow, or both.
Triggering a DAG externally fails consistently for 1% of requests, and 100% of the time when the web server API goes down, which is too often (502). I understand that in the realm of Airflow, triggering a DAG externally is perhaps a bit of an anti-pattern, but it should not have been documented then:

https://cloud.google.com/composer/docs/how-to/using/triggeri...

I have cases open with the composer team, and I am upgrading the cluster again this weekend, but I have lost hope at this point; I am designing a solution to stop using external triggers.

The AWS and Azure numbers are actual revenue, but the Google cloud number is a ‘run rate’, meaning, pick a particular high point, call that steady-state, and multiply accordingly to get an annual rate.

I am not doubting Google’s success with GCP. However, they should provide the actual revenue, in addition to the run-rate or whatever other vanity metric they want. Not doing so just makes me wonder about their lack of candor.

I think it's more complicated than that (for all three vendors) because customer contracting can involve all sorts of convoluted terms, especially mega-deals. Using ARR makes more logical sense because it provides a less messy way of accounting for things like guaranteed future consumption commits.
Interesting, so it's really just a means of smoothing then? Do all three use run rate as their key metric then?
As a public company you wouldn’t exaggerate it that way. An $8bn run rate means $2bn last quarter. We often look at annual numbers so reporting a run rate makes sense. It’s not meant as a deception.

Anyway run rate doesn’t project growth so if anything, it’s understating the numbers.

That sounds great- congrats on the impressive growth last quarter.

However, it would be more insightful if the earnings report made it clear that the revenue was actually $ 2 BB last quarter in addition to giving a 'run rate.' We would then be able to apply an apples-to-apples comparison with AWS, which made $8.4 BB last quarter, growing 37% from the quarter last year, with a 'steady-state run rate'(no growth) of $33B.

That seems straightforward, don't you think? Giving partial info when compared to your peers will only encourage deeper scrutiny.

As for the point about ARR in the other comment, it would be good to see how much of the revenue is due to pre-existing commitments vs. on-demand use, but you won't get that just by using a single 'run-rate' number.

I've been in the cloud space for a relatively long time, worked at AWS 2008-2014, and have followed all three big cloud providers since inception.

My personal view is that both Azure and GCP boost their "cloud" revenues by 50-60% - e.g. GSuite is included in GCP, some various Office and Enterprise software products by Microsoft are included in Azure.

They do this primarily to convince corporate customers that they've closed the gap with AWS, and their service is now comparable.

From a technical standpoint, and still IMHO, Google has fewer products, but several of the ones already available are usually quite superior to AWS. AWS has of course the broadest set of services. Azure is still quite a mess, but that doesn't prevent a thriving ecosystem, an easier integration, etc, to convince many customers to use Azure.

It's going to be a long battle, for a huge and growing market. And unfortunately for their customers, it means that lock-in, FUD, and other nefarious practices will be common.

You could flip that and say that AWS under-performs in straight SaaS like github, gsuite etc.
Even if your numbers are correct, it is still most relevant to customers that azure can scale up to <some big number>.

That part of that big number is used internally is not an issue, dogfooding is a good thing and further demonstrates that cloud is a real priority for Microsoft.

So what? Had gsuite/office been made by a 3rd party company, you would have counted them in, right? And had Google/MS not have a cloud presence so these products would have been hosted in a different cloud, they would have been counted in your estimation as well, right? This case is technically no different than the others.

It is a fact Office/GSuite are rather reliable and fast. It's a credit to their respective clouds, and obviously propels them forwards in their competition with AWS. The respective companies are right to include these in total revenue/marketshare, and it is good to see that they eat their own dogfood. We could have flipped that around and ask why Amazon doesn't trust AWS, but I recall that Amazon does have a few things there too...

Is the cost of GCP/Azure also deducted from GSuite/Office profits?
I suspect it is, though I haven't delved in their financials to prove it. Giving a different internal pricing to their own product could be constructed as an anti-competitive action, and someone would have sued them for that. If they merely didn't report it [that is, didn't count it in Office/GSuite costs], it would sound like hosting for free which again could be considered anti-competitive, so that sounds unlikely to me.

My original comment was just saying that there's no good reason not to count GSuite/Office in respective GCP/Azure marketshare. Obviously these cloud-products could have been hosted in different clouds, but they aren't.

I disagree with you.

If you want to tell me how big your public cloud is, in terms of revenues, you are not going to count revenues that come from software than runs on top of it.

By that account, AWS could say: 30B, plus 200B more, taking into account the revenues of Salesforce, Elastic, etc.

It makes no sense to conflate these two very different categories.

*"revenues that come from software than runs on top of it"

I am not sure I understand you here. Do you mean "what the software is paying to run on the cloud" (even if the payment is 'merely' internal accounting) or "what customers are paying for the software that runs on the cloud"?

I meant the former, and that's the way I read your original comment. But your latest comment seems to refer to the latter? I agree these would be different categories (IaaS/PasS vs SaaS).

I was saying that MS/Google should be able to count Office/GSuite hosting expense on their cloud - it would be no different if it were some 3rd party product - not that they could count what customers are paying for these products.

I wonder how much Google's deeper integration with kubernetes and the influence of pushing the industry towards using k8s is going to help them in the long run. Could be pocket change at this point but as newer customers come onboard it would be interesting to see if this helps their numbers more.
Google's strategy of standardizing microservice infrastructure around kubernetes is likely to benefit them greatly in the long run. Any large company wants to avoid vendor lock-in, and [kubernetes + containers] offers a compelling alternative
Disingenuous mix of per-quarter and per-year figures.
TIL: The "run rate" is extrapolating a year's revenue from some shorter period. So the headline means, "Google Cloud earned $2 billion last quarter, which will be $8 billion a year if we can keep it up."

So it makes the numbers more comparable with "per year" as the standard rate, but with all the obvious hazards of extrapolation.