Didn't realize Google Cloud was so far behind in revenue and growth. Seems like it would be risky for companies to switch to Google Cloud because it might end up in the graveyard.
I understand the narrative that's going to carry after that article the other day, but I've really enjoyed working with GCP over the last 2 years.
The structure for VM use and their Live Migrate capability for upgrading the hardware underneath with no downtime has been excellent on my more long-lived servers.
My biggest wish is some equivalent of Aurora for PostgreSQL.
Another fairly happy GCP user. They have their issues (everyone does), but GKE is pretty amazing if you don’t have the resources for an in house k8s on AWS team.
I am however fairly pessimistic about their chances. Especially considering the person at the helm is ex-Oracle VP.
I also don’t understand their alternatives... do they want AWS to own the cloud market? They’ve operated Android as the bulwark against Apple, why not view GCP in a similar light?
> They’ve operated Android as the bulwark against Apple, why not view GCP in a similar light?
Android has a lot more synergies with their consumer ads business. It's the extreme version of writing your own browser.
GCP is completely different; it's in the B2B space, and it's a pretty standard business relationship between businesses and a supplier. Being customer-focused and business-focused isn't in Google's DNA.
On GKE, you don't need to define node-pools if you tick enable node-autoprovisioning. This is effectively an EKS fargate profile without any of the limitations (discrete node size, inability to mount volumes etc)
On GKE there's no need for bastion hosts, you can connect to private nodes automatically by tunneling through identity-aware-proxy
On GKE, there's no faffing about with aws-auth-configmap, nodes automatically join the cluster
On GKE, you can use regional persistent disks if you need to store state rather than single zone EBS
On GKE, you can use a single static ip for your nginx-ingress load balancer without needing to faff about with AWS global accelerator (which in any case gives you multiple ips)
On GKE, managed node pools are automatically repaired and upgraded with the choice of google's OS or ubuntu unlike Amazon Linux which is currently the only choice for EKS' managed node pools. GKE can automatically use spot instances unlike the farce on AWS with a 3rd party (spotinst) charging a premium for the same functionality
On GKE, managed istio is ticking a box versus a self-install on EKS
On GKE you need to worry much less about ip exhaustion as they use alias ips vs dedicated ENIs for EKS
Associating IAMs with kuberenetes service accounts is much easier with GKE workload identity than with EKS' oidc webhook
GKE has several features which EKS doesn't (calico, vertical pod autoscaling, binary authorisation, export of cluster data to bigquery)
GCP is fantastic! I've been especially pleased with their network (the GLB is awesome!), BigQuery, Bigtable, GCE live migration, and of course... GKE. I cringe at the thought of having to run Kubernetes myself or settle for an inferior managed solution from Azure or AWS.
> Seems like it would be risky for companies to switch to Google Cloud because it might end up in the graveyard
It's sort of a Catch-22 for them. They are this far behind in part because people fear they might not be around long term, and they might not be, because people don't want to sign up, because they fear it might not be. :)
To some extent they are doing this by embracing Kubernetes. For us, almost all of our infrastructure other than databases is running directly on GKE, so moving to AWS or Azure would be fairly easy.
Chasing AWS's APIs probably isn't a good use of their time, and wouldn't encourage a lot of people to move.
API compatibility is not the biggest blocker to changing clouds -- moving one's data is.
The could work out a deal where outbound data from Google to AWS is free, but that probably wouldn't be in their best interests either.
I think their strategy of being the best place to run K8s is a good one. If you standardize on K8s you can run on any cloud, and they want to be the best place to do that.
You’re falling for the “OMG Lock In” fallacy. Even if all of the APIs are compatible, if you have any type of scale and even if you’re just running a bunch of VMs, the pain of migrating infrastructure, security, data, network, auditing, connecting back to your on prem network is often a painful multi month process with huge risks of regressions
I worked for two shops that migrated to GCP, one from leased space in DCs, one from AWS.
> Seems like it would be risky for companies to switch to Google Cloud because it might end up in the graveyard.
Yeah, there's definitely that risk, and in one of those cases, I disagreed with the move because of it.
That said, depending on what you're doing, GCP's key-value store (BigTable) and analytical DB (BigQuery) struck me as technically more sound than the AWS versions (DynamoDB, Redshift/Athena). As an ecosystem, AWS was far superior, there are more offerings, and they tend to have more features. It's just that they're most likely a managed version of an open-source project.
One place moved specifically because of BigTable. They were on HBase (and not happy with it), so it was the least painful migration option.
Google is unparalleled at building efficient data centers. Cloud already has tons of enterprise customers with contracts. They've spent $$$$$$ creating the infrastructure to be able to cash in over the long term.
Anyone who thinks Google will get out of Cloud (already more than an $8 billion business [1]), as if it were comparable to Reader (which made $0, AFAIK), isn't looking at the economic realities here.
I repeat: Google Cloud has more than $8 billion in annual revenue. And AWS proves that clouds are a profitable business.
I just can't understand how anyone can suggest with a straight face that Google might shut it down.
Corporate leaders set stretch goals all the time -- that's part of their job. Not meeting them just means certain VP's might not get their full million-dollar bonuses -- not that a billions-dollar business will shutter and lay off 1000's of people.
In a large company, sure, but in Google? They nuke their own stuff all the time, users be damned. Since AWS offers a larger ecosystem, and doesn't have the Google product TTL of like 5 years, I really don't know why anyone would use GCP unless they're going for multi-provider redundancy
Tired and bad meme is tired and bad. Why would prior evidence of Google shutting down things that are not anything like GCP suggest that Google would shut down things that are like GCP?
There wasn't, i.e. the thread existed, but the purported discussion of shutting down GCP did not occur as written. But don't take my word for it, take the word of Corey Quinn, the AWS consultant who originally invented the meme that Google might someday shut down GCP.
Most of the google products that have been shut down that HN loves to harp about were shut down not because they failed to reach some sort of goal, but because they had no goals at all. They are by and large the work of a few people and nobody felt like maintaining them. The one exception is Google+, which notably unlike Google Cloud didn't require spending billions of dollars on datacenters and undersea cables, or on hiring/diverting thousands and thousands of engineers. Nor did it require building a third-party partner ecosystem and signing long term contracts with deprecation windows. Nor did it have a projected total market revenue larger than all digital advertising.
Well, I know people who work there who’ve told me they stopped hiring for GCP and are also letting people go from that group. Also I’ve heard it mentioned maybe unofficially that Google themselves said they would shut it down if they weren’t #1 by a certain date.
Highlights from fiscal year 2019 compared with fiscal year 2018 included:
Commercial cloud revenue, which includes Microsoft Office 365 Commercial, Microsoft Azure, the commercial portion of LinkedIn, Microsoft Dynamics 365, and other commercial cloud properties, increased 43% to $38.1 billion.
You’ve just kind of described the problem with GCP. It doesn’t matter how technically good they are if they don’t have good Enterprise support and they have no ability to meet the enterprise customer where they are.
The first step of a major cloud migration is often a combination of hybrid solutions and lift and shifts.
Revenue is inconsequential. What is there profit? Cloud is profitable for AWS and MS. We don’t know what the margins are for GCP.
I don't know what you're talking about. By all accounts, GCP enterprise support is perfectly fine. Why do you think it isn't? Again, it's not consumer Gmail or YouTube accounts. Major companies wouldn't be signing contracts with them if they didn't.
And in what universe is revenue inconsequential? Profit comes whenever you want to stop investing in growing. There is zero reason to believe that Google will be unable to turn a profit off GCP whenever it decides to. Particularly given that it was building datacenters for Search before AWS even existed.
It’s not really about “will they shut it down?” But rather, will Google continue to invest heavily in GCP if it’s not gaining traction at the pace they want it to? Maybe they’ll find better ways to spend that money. Microsoft got out of the mobile phone business, which seems ludicrous to me, but they weren’t gaining traction, so...
Google really dropped the ball on missing cloud computing. With their expertise, they should have owned the space, it's just never the sort of idea that would bubble up in that org.
But balloons to deliver internet in rural areas? They're the leader in that space.
I think it's more they wanted to do it on their own terms. They wanted to provide the best most scalable cloud platform ever in GAE, but users wanted more of a dumbed-down lift-and-shift-friendly variant that AWS offered.
And what if some Google employees decide they don’t like your organization? Google has already demonstrated they will cave in to that kind of pressure. How can any non-trivial sized organization bet the farm on gcp after that?
My biggest takeaway from this article was that while Google publicly reports its "cloud" revenue (including GSuite revenue) as $8 billion a year, its cloud computing revenue (the portion that competes with AWS and Azure) is at $1-2 billion a year.
That would explain why Google would be willing to shut down Google Cloud in 2023 if they are not one of the top 2 players in this space. (Additional context and previous discussion below.)
Doesn't Microsoft do the same thing? The last revenue numbers I remember seeing from them were for some category called "Intelligent Cloud", which includes:
* Server products and cloud services, including Microsoft SQL Server, Windows Server, Visual Studio, System Center, and related CALs, Microsoft Azure, and GitHub.
* Enterprise Services, including Premier Support Services and Microsoft Consulting Services.
>That would explain why Google would be willing to shut down Google Cloud in 2023 if they are not one of the top 2 players in this space.
I wonder how much of an effect this disclosure will affect the outcome?
I've used AWS heavily for work-related services and have touched GCP a few times for little side projects in the past. Knowing that they'd be prepared to shut down GCP if they're not market leader or market second within a small amount of time ensures I personally will never, _ever_ use them for enterprise work, and I'm sure I'm not alone in that.
I’m in the middle of a painful migration from AWS to GCP and my advice is definitely stay away from GCP.
The cloud tools do not usually have real parity or interop with other cloud tools as claimed (eg interop with boto or s3), many gcloud utilities have way worse performance and memory issues, the docs are plastered with up-selling ads and, inconsistent service account auth vs IAM-style user auth issues, and on top of all that, there is this risk Google shuts it down.
Even if they don’t shut down GCS or GKE core stuff, there are so many peripheral products like managed big data solutions, Cloud Functions, etc., that seem like _classic_ cases where Google will abruptly shut them down.
Those resources should not be read as implying that there is any chance GCloud gets "shut down" in 2023 or 2033 or 2043 or anytime before the recognizable end of computing as a business. There are issues, to be sure, but this is not Reader.
Google maintains strategic leadership in infrastructure to efficiently run Ads. The ultimate addressable market in the cloud business is as large or larger than the ads business.
Cloud also is just getting into the middle game, where commodity storage/compute services start to give way to specialized and domain specific storage/compute/analytics/intelligence. Google has unique strengths there, and a number of opportunities/advantages vis a vis Microsoft/AWS.
But rather than dominating in Cloud, Google treats it as a hobby- like TV was for Apple for about a decade. Google can perfectly well continue to treat it like a hobby. They are well ahead technically, and have plenty of cash to maintain that leadership underneath Ads.
Those resources are reporting on the framing that- for the right leaders inside Google, there is an opportunity to turn their infrastructure capabilities from a strategic skillset into a leading business. That's all.
Maybe they will, maybe not. But even if they don't, there will be no "shutting down" on any near term time frame. Even as a hobby Cloud will be big enough to be alive for a long time.
Nice how "theinformation" dumped this "article" out on dec 17 when, you know, everybody is going on vacation. Some of us are back now (as it turns out, I work in GCP, but I code).
I suspect there will be an official response to the response to this pile of crap at some point.
And speaking of crap, how do you get from "risk loss of funding" to "shut down"?
This is nonsense, based on ignorance/misunderstanding blown out of proportion, compounded by a popular meme that's not actually applicable. But don't take my word for it, take the word of Corey Quinn, an AWS consultant who (I believe, maybe someone knows an earlier example) originally popularized the joke that Google might someday shut down GCP, and whose current dominant refrain is that Kubernetes will eventually fail and be abandoned.[0]
> That would explain why Google would be willing to shut down Google Cloud in 2023 if they are not one of the top 2 players in this space. (Additional context and previous discussion below.)
I don't see anything mentioning that Google will shut down GCP unit in 2023 - I see something saying that GCP unit will risk losing funding - presumably this is not quite the same thing. Losing funding to me would mean they will have to fund themselves on their own revenue.
It's actually a great testament to Microsoft's execution that they were able to catch up (to some extent at least), despite AWS having a seven year head start.
And this is against Amazon, which is known for execution. I suppose we're in for a fun couple of years, especially with the JEDI contract soap opera and all that.
Great observation, with a very minor caveat: when Gates & Allen started out in programming languages, they were an awesome first mover. I highly recommend the "Hard Drive" book which goes into a bit more detail.
However, in everything after that very first stage they were an amazing second mover: DOS, Word, Excel, etc. And now Azure.
Microsoft's great advantage is enterprise bundling. Not only do they already have sales relationships with every big IT shop in the world - they're selling both software and cloud (which means they can discount one or the other).
To Microsoft's credit they maximized this advantage very well.
Id argue Microsofts strong suit here is making themselves look competitive. During the web server wars Apache reigned supreme, until the numbers shifted in favour of IIS which overtook Apache for a time. Was this a better product Apache was fending off? no. Microsoft had paid hosting providers to park web domains and registrations on IIS servers so Netcraft would finger them as #1.
now we have Azure and its the same thing all over again. Azure Cloud has lots of tenancy, but these numbers I suspect are bogus. Azure runs Office 365, Exchange, Bing, and Xbox Live assets as well, and these are all "tenants." MS also bundles discounts for business critical things like desktop and server licenses into accepting Azure credits, so its likely many of these tenants are on paper only.
I don't think we can tell for sure before they start releasing numbers for cloud proper. Both Azure and GCP are problematic in that regard (see the acrobatics that Stratechery was going through while trying to deduce GCP numbers).
That being said, they've been closing the feature gap, and have shown themselves able to pick up large contracts (controversy over JEDI notwitstanding), so the market certainly seems more competitive than a couple of years ago.
As for factoring discounts etc. into cloud numbers, I confess that I don't have a good answer.
Bits and bobs run on top of pure azure, and some infra is definitely shared, but anything cloudy that came up around the same time as or before azure (like Bing and o365) used and for the most part still uses their own data centers and deployment tech.
O365 workloads are distributed across a number of environments. While Exchange mostly lives one its own hardware, SPO is very distributed now - it still has its own hardware, but also has a lot of data and services that run in Azure. And O365 is very dependent on AAD.
My recollection is that IIS marketshare started falling when domain parking became a thing. Particularly once the registrars got involved.
IIS was always a good web server. Classic ASP had unique ways of shooting yourself in the foot... but so does everything else. And in those olden days a retail box of NT Server barely cost more than a Stronghold license.
I think Microsoft leveraged their Office lock-in, additionally to luring some users with apps like one-note/one-drive.
I use Azure outside the US from central Europe and the performance of their cloud services is just terrible.
Personally, I don't really like the move to the cloud at all. The whole MS palette of services from the OS to Office doesn't really convince me.
I am surprised the cloud market is actually that small. Especially since classic hosting is just included. And a growth of ~250% compared to last decade? A decade is pretty long...
Our business "jumped" into the cloud because of the reliance on MS office products. It was not done out of love.
Using market share along isn't a very good measure, because it hides that fact that they entire market is growing significantly, far outpacing the market share loss.
It's not that Microsoft is stealing customers from Amazon. It's that both companies are brining in lots of new to the cloud customers, and Microsoft is just doing it faster.
But both businesses are growing at enviable rates.
As a hypothetical CTO worried about the political consequences of failure when choosing a cloud provider, why would I choose GCP? If I choose AWS and things go wrong , I made the safe choice “and no one ever got fired for choosing AWS”. If I’m already running a MS shop, Azure is a safe political choice. Every time something goes wrong, even if it could have happen with any provider, you will always be questioned about why you chose GCP.
Besides I’m sure that AWS and Azure are some where in the right top corner of Gartner’s magic quadrant.
Typically GCP is chosen for up front volume discounts and bundling deals with GSuite and other tools, even more so for companies that offer GSuite-as-service accounts within other products, like business tools or hosting platforms.
The problem is Google can’t make up the market share from only pressuring “lackey” companies that it has leverage over to use their cloud, it has to have a big scale draw for otherwise unaffiliated customers, and to a lesser extent can’t sustain these initial volume & bundling discounts indefinitely.
There is no special reason that MS workloads that should be on Azure. AWS is always bragging that they have twice as many Windows VMs as Azure. The only two features that I’m aware of on AWS that won’t work for Windows is lambda and Fargate.
AWS SDKs, documentation and support for Windows/.Net Framework is excellent.
I’m not making any technical arguments against Azure. I haven’t used it besides what is now “Azure Devops”.
As far as Kubernetes. It’s inconsequential when a major corporation is considering which cloud provider to use. Most existing workloads on prem are not using k8s. That goes back to knowing how to meet the customers where they are and knowing how to do a “lift and shift” as phase 1.
I'm not saying they are technical arguments. I'm saying they are the same acceptable answers to "why did you choose (not AWS)". Because the platform I chose is also part of the foundation of the tech stack we run on.
Kubernetes - which every cloud provider now offers as a managed service - might win you a few startups but won’t win you many major enterprise companies.
You'd probably be surprised. Kubernetes is a hot enough buzzword that even companies as bizarre as Chick-fil-A are getting on board, with a hilarious architecture.
Edit: Way past the getting-on-board stage, apparently. It had slipped my mind that the article I linked is from 18 months ago and already described an extensive rollout.
FYI I saw a presentation by the team at CFA after that blog post was written. They said the in-store K8s deployments were only running greenfield, non-production applications. I don't know if that's changed.
That’s a great anecdote, but small fry when compared to all of the other F500 companies that MS and AWS are able to court by using their thousands of partners that are intimately familiar with enterprise on prem workloads and know how to slowly migrate customers.
MS has decades of building relationships with large companies that it can leverage.
If Google’s only proposition is that it’s cheaper, any AWS Solution Architect can negotiate lower prices from AWS if they want to secure a client that’s large enough.
How do you think major enterprise customers are formed?
Also GCP already does have some large enterprise clients. Some of them have had a symbiotic relationship with them, working closely to create the features they need as they scale. That kind of treatment is hard to get with AWS where most enterprises would be small fry by comparison.
There are also enterprises that don’t want to be on Amazons stuff for reasons.
How many hot startups have become “major corporations” in the last 5 years.
All you have to do is watch the thousand plus reinvent videos to know that AWS does work with many major companies. They also have an extensive partner network.
Microsoft is breaking apart what WAS Windows Server, into discrete managed services. Identity Services, File Services, etc. If you have a need for what was formerly Windows Server provided stuff, and youre rebuilding infrastructure, you would get a lot of compatibility from Azure. Then, as you need other compute and storage resources, youre building on top of a service contract you already have, instead of spreading your data all over the clouds.
As is Amazon. They have hosted Active Directory and hosted Windows File Servers. Amazon is not crazy enough not to cater to Windows workloads when that’s what many enterprise companies are using.
But amazon doesnt make Active Directory. Microsoft can continue to create new versions and new features that are released as a service, but not in the on prem component. You already see that in exchange, sharepoint, lync.
Because as a CTO you usually answer to outcomes not defensibility. If you frequently make defensible decisions and they keep not panning out, your CEO will fire you.
Yes, but as a newly hired CTO, you want a win...and failing that, a loss you can defend to a non-technical CEO. "Industry leader" or "we already are in bed with them" is a reason the non-technical CEO can understand.
> If you frequently make defensible decisions and they keep not panning out, your CEO will fire you.
That could be a case for startups but it is certainly not the case for established companies.
The job of a CTO of an established company is not piloting a formula 1 car. Rather, it is to prevent reckless technical decisions from taking an ocean liner and smashing it into an iceberg.
Google's Cloud TPU makes AI training like 10x cheaper than the cheapest GPU instances that AWS and Azure offer. Plus Cloud TPU is perfectly integrated into TensorFlow, due to both being Google products.
In my case it was 10x. We're training networks to estimate optical flow on Sintel and KITTI using TensorFlow. On GPU, we hit the 12GB limit which caused TensorFlow to log a message that convolution performance would be limited due to not having enough cache RAM. On TPU, we didn't hit the RAM limit.
Oh and amazingly enough, I took the exact same Python file that I built on GPU on my workstation, ran it in Google Collab with a TPU connected, and I really only had to change the configuration to make it work. There was no need to change or tune anything. I was pleasantly surprised by how well that worked.
And let me rant a bit longer, when we started this project I actually first tried to use AWS because our company already had billing set up with them. But AWS was using CUDA 10.0 for SageMaker, which has a known crash- and nan-inducing bug for 2D convolution. AWS support being non-existent as always, we couldn't make things work on AWS and then went bare metal for our GPU training with hetzner.com - which simultaneously also shaved like $30,000 monthly off our bill. AWS's EC2 GPU instances are really expensive when compared to a dedicated root server.
I’m not arguing either way. But that’s still not where the money or growth is. Cloud computing only makes up 5% of enterprise (big Fortune 500 companies) computer spend.
MS recognized moving enterprises with boring workloads to Azure was where the money was along with hybrid computing. That’s how it was able to succeed so quickly.
Besides,Microsoft has always been focused on the dark matter developers doing boring stuff. Their tooling is far better than Google’s and their IDEs that developers are already using integrate with their cloud offerings.
From watching NVIDIA's presentations, it sure sounds like AI is growing exponentially. Plus Google needs AI to advance for Waymo to become safe and profitable. Looking at other companies in my area, all of them have plans for AI, too, even if it is trivial stuff like automating the most repetitive 5% of support inquiries. So if Google can beat AWS out of the AI waters early on, that might become very valuable in a few years.
I may be wrong, but I recall reading Azure grew quickly because they already had MS contracts in place with many enterprise companies. They simply said, we’ll also hold your hand in the entire cloud process.
This was critical because by then if you weren’t on AWS, you probably didn’t have the DevOps team to do so - MS was now willing to lend you enough consultants to do it for you.
The number of spend shouldn’t be astronomically high. Most workflows in cloud computing are commoditized low-margin. It also explains why Google wants to do high-margin AI stuff, and why AWS is happy with low-margin areas.
That’s kind of my point and what I’ve been saying. Microsoft doesn’t tell a company that they should be using their new cool technology because we are “smart people” (tm) who know what’s best.
Microsoft would say:
“If you have a workload that still requires Windows 2000 and SQL Server 2003. We will help you migrate it to Azure as is and we will extend your support contract.”
While Google will talk about migrating your entire workload to K8s.
Plus, Microsoft also has K8s support, so not only can they can "lift and shift" the existing dark matter applications, but they can also offer a conversion path/greenfield path.
Sure, but it adds an illusion of portability, which is partly what GCP is taking advantage of in focusing it and definitely what Microsoft here is taking advantage of in promoting K8s to dark matter developers:
Part of Microsoft's implied message in heavily marketing K8s is that they know GCP is heavily K8s-based, and know GCP is heavily competing with them on price, and they are counting on some of that "dark matter conversion work" wanting to move towards the "more portable solution" in order to "maybe when the moons are right eventually shift over to GCP or something cheaper than Azure". By offering that conversion path at all, and knowing Enterprise's weakness for sunk costs development over from-scratch rewrites, they are playing a very different game than what GCP can offer.
Portability is always an illusion once you have any type of scale. Everyone who has ever done any type of large implementation or migration knows that.
Even if you did everything in a “cloud agnostic” way, once you have even a couple of dozen VMs, your networking infrastructure and security rules, your user permissions, your DNS entries, your site to site VPNs, your client to site VPNs, all your data in databases, security and compliance audits etc. migrating your infrastructure is going to be a months long process, involving project plans, regression tests, and still have little to show for it.
I was involved in a much smaller migration as a dev lead when all we had to do was migrate to WorkDay for HR, payroll, and time tracking. That involved tons of integration work and hiring Workday consultants. Not to mention consolidating all of the other SAAS solutions that we had to integrate when we merged with another company.
Which is exactly part of my point in how shrewd Microsoft's marketing is here. K8s exactly has the marketing "smell" of "cloud agnostic", without being in fact provably so. Microsoft can compete with Google better in that arena than Google can precisely because they have other services. It's a "yes, plus" situation. "Yes, we do K8s too, but also we can lift and shift your existing operation in the meantime while you build it." That has a lot of hidden lock-in power and I presume Microsoft knows that very well, even better because they know after that "lift and shift" many dark matter companies will just stop there because while they could migrate, migrations aren't free or cheap and they can kick that can down the road possibly forever. Managers every day are persuaded by the possibility that they could migrate to K8s if they wanted to, whether or not that is ever the real plan. It's why "lift and shift" really is a fascinating game changer that Azure has that neither GCP or AWS can quite so directly compete with.
It's why Azure's K8s strategy is so fascinating: precisely because the more they make it "cloud agnostic" on paper and are good K8s/Docker citizens the greater power they have for Manager mentality lock-in to the non-cloud agnostic stuff (especially "lift and shift").
GCP has a really strong ecosystem around AI. If you end up dealing with some type of ML problem with absurd memory requirements, TPUs can be useful there. A lot of recent Transformer architectures have these requirements.
Also, a huge amount of cloud costs are based on egress, so having it on one cloud saves you resources.
All the major cloud storage providers tout geo-replication so that if an earthquake or meteor takes out a datacenter there’s still other copies.
But nothing can protect against an administrative SNAFU where you - or your cloud provider - and accidentally or intentionally deletes your account and everything’s gone in an instant (yes, there’s usually recourse and backups to recover from - but that’s hours or days of downtime).
Sometimes it isn’t even the cloud provider’s fault: see Adobe’s blameshifting onto the US executive for their dropping all their Venezuelan customers.
While Azure has the “Black Forest” unit which is legally - and technologically - firewalled off the main Azure cloud - and you can also license Azure Stack for running many (most?) of Azure’s platform on-prem or self-hosted, I haven’t seen similar offerings from AWS or Google.
GCP also considers it copyright infrigement if you use their AI services for something actually useful, like, for example, self-driving car software. So it is better to stay away from it, unless you only use it for fun.
Note that MS might not do it faster in real numbers, just that percentage wise it's easier to grow when you're small. 15% of 20B is still a lot more than 30% of 2B (out whatever the real numbers are).
Likewise, we run everything in DO and it's around 30% less so far, but we're also able to afford managed databases at that rate. To make them redundant will bring us closer to our AWS rates, but... I'm very happy to have managed services at the moment, and the value is felt both in productivity and finances. I'm really happy we moved over.
An "order of magnitude superior" is generally the minimum that causes me to start expending engineering effort to move from something implemented and nominally working to something unfamiliar and unimplemented.
10% savings is almost never worth engineering effort. Those engineers are probably worth more than 10% by implementing features or fixing a bug for a customer.
However, even if 10% savings is $1 million, the edge odds probably still aren't in your favor. You may burn engineering time and the project fails or takes 3 times as long. Those engineers are effectively "dead" during that time and whatever contribution they could have made elsewhere is lost. For some reason, nobody ever factors that into their "savings projections".
Obviously, if I've reached the point that 10% savings is $10 million, I'm in a different bucket. But having $100 million in cloud spend probably indicates a lot more problems than AWS vs GCP.
Am using it at current client. It's very good. GKE is absolutely brilliant, and some of the other stuff is also great.
On the other hand, some stuff is not quite there yet. The metadata server to handle instance/pod IAM (i.e. Workload Identity) is wobbly at best. To be fair, it's in beta - but then most of their offerings seem to be - which then means you can't rely on it for business-critical applications since it may wobble at the wrong time and there's no SLA.
This is just one example off the top of my head - I'm on holiday and not making massive mental effort to remember work. :)
AWS is still my preferred cloud provider. Their APIs can be daunting at times and you have to build off many lower level concepts until you get something working but when you do it tends to be rock solid - and under your control. I prefer their approach of giving you the best lowest level services to build on, and then building on top of those themselves for those who want more abstraction than the Google way of "we're building something you'll want to use - we'll tell you when it's available but it will probably be another year or two until it's GA and covered by any SLA."
One thing to note is that the cloud is only about 5% of total IT expenditures. So getting data which has never been on the cloud is more important than taking data away from competitors.
Here's the thing though, they actually don't have to fend off rivals. Amazon's cloud business is based on excess capacity that they have as one of the internet's biggest chokepoint in traffic (post-Thanksgiving holiday spending). When I took managerial accounting classes, we looked at situations where it would seem on paper that producing a certain good would lose money, but when you compared total profitability in letting excess capacity sit idle and distributing fixed costs like rent and overhead, you can actually sell a good for a loss. Amazon doesn't have to make more money than they spend on cloud computing, they only have to make more money than what they would have had to spend on cloud computing without competing on the market.
Financial inquire: I would like to see the breakout from MSFT on what is actual Azure revenue. That would be a good comparison with AWS on the revenue side. MSFT now includes the O365 entire suite and other SAAS products with their cloud revenue. Office is huge alone.
Also, genius tactic, MSFT gives you free SQL Server, Windows Server, and Exchange when you buy Azure credits. I have friends with expiring Azure credits coming up soon, waiting to see how MSFT handles. MSFT has massive cash, so prob extend for free or some other magic move.
Edit: For context, these are multi millions dollars in Azure credits, some 20M+, not being used.
So, there's a thing I wonder about cloud businesses, Amazon or anyone else's. How much of this is funded by businesses that are not profitable, but are running on VC money? I recall the IPO documentation for companies like Uber stating a stupendous amount of their $$ going for AWS.
I am reminded of Paul Graham's description of Yahoo's advertising business in 1999. Lots of paying customers, things seem great, you're solidly profitable. But what if most of your customers aren't profitable? What happens when the next recession comes?
Not saying cloud computing is going away, but I think its current apparent growth rates are being propped up by a lot of Softbank (and similarly unprofitable) investments, which will go away fast when the next recession comes. Until then, we don't really know how profitable cloud computing is as a business. It may not be as good as it looks right now.
138 comments
[ 0.21 ms ] story [ 178 ms ] threadThe structure for VM use and their Live Migrate capability for upgrading the hardware underneath with no downtime has been excellent on my more long-lived servers.
My biggest wish is some equivalent of Aurora for PostgreSQL.
I am however fairly pessimistic about their chances. Especially considering the person at the helm is ex-Oracle VP.
I also don’t understand their alternatives... do they want AWS to own the cloud market? They’ve operated Android as the bulwark against Apple, why not view GCP in a similar light?
Android has a lot more synergies with their consumer ads business. It's the extreme version of writing your own browser.
GCP is completely different; it's in the B2B space, and it's a pretty standard business relationship between businesses and a supplier. Being customer-focused and business-focused isn't in Google's DNA.
It's sort of a Catch-22 for them. They are this far behind in part because people fear they might not be around long term, and they might not be, because people don't want to sign up, because they fear it might not be. :)
They could still make money by being far more efficient when running their software natively on their own stack.
API compatibility is not the biggest blocker to changing clouds -- moving one's data is.
The could work out a deal where outbound data from Google to AWS is free, but that probably wouldn't be in their best interests either.
I think their strategy of being the best place to run K8s is a good one. If you standardize on K8s you can run on any cloud, and they want to be the best place to do that.
Still have the data problem though.
> Seems like it would be risky for companies to switch to Google Cloud because it might end up in the graveyard.
Yeah, there's definitely that risk, and in one of those cases, I disagreed with the move because of it.
That said, depending on what you're doing, GCP's key-value store (BigTable) and analytical DB (BigQuery) struck me as technically more sound than the AWS versions (DynamoDB, Redshift/Athena). As an ecosystem, AWS was far superior, there are more offerings, and they tend to have more features. It's just that they're most likely a managed version of an open-source project.
One place moved specifically because of BigTable. They were on HBase (and not happy with it), so it was the least painful migration option.
Anyone who thinks Google will get out of Cloud (already more than an $8 billion business [1]), as if it were comparable to Reader (which made $0, AFAIK), isn't looking at the economic realities here.
I repeat: Google Cloud has more than $8 billion in annual revenue. And AWS proves that clouds are a profitable business.
I just can't understand how anyone can suggest with a straight face that Google might shut it down.
[1] https://www.sdxcentral.com/articles/news/google-cloud-annual...
Corporate leaders set stretch goals all the time -- that's part of their job. Not meeting them just means certain VP's might not get their full million-dollar bonuses -- not that a billions-dollar business will shutter and lay off 1000's of people.
It is all their own fault though I think: They worked long and hard to build that reputation.
https://www.lastweekinaws.com/blog/google-cloud-is-probably-...
Commercial cloud revenue, which includes Microsoft Office 365 Commercial, Microsoft Azure, the commercial portion of LinkedIn, Microsoft Dynamics 365, and other commercial cloud properties, increased 43% to $38.1 billion.
https://www.microsoft.com/investor/reports/ar19/index.html
The first step of a major cloud migration is often a combination of hybrid solutions and lift and shifts.
Revenue is inconsequential. What is there profit? Cloud is profitable for AWS and MS. We don’t know what the margins are for GCP.
And in what universe is revenue inconsequential? Profit comes whenever you want to stop investing in growing. There is zero reason to believe that Google will be unable to turn a profit off GCP whenever it decides to. Particularly given that it was building datacenters for Search before AWS even existed.
GCP and Google are completely different. Nothing except a few inconsequential internal Google apps run on GCP.
Building servers and knowing how to address the needs of the enterprise are completely orthogonal.
Sorry the only citation I have is a podcast.
https://overcast.fm/+RWDUSvtD8
But balloons to deliver internet in rural areas? They're the leader in that space.
That would explain why Google would be willing to shut down Google Cloud in 2023 if they are not one of the top 2 players in this space. (Additional context and previous discussion below.)
[1] The Information: https://www.theinformation.com/articles/google-brass-set-202...
[2] Discussed on Hacker News previously: https://news.ycombinator.com/item?id=21815260
* Server products and cloud services, including Microsoft SQL Server, Windows Server, Visual Studio, System Center, and related CALs, Microsoft Azure, and GitHub.
* Enterprise Services, including Premier Support Services and Microsoft Consulting Services.
(https://www.microsoft.com/en-us/Investor/earnings/FY-2019-Q3...)
I didn't think the cloud revenue split was public knowledge.
I wonder how much of an effect this disclosure will affect the outcome?
I've used AWS heavily for work-related services and have touched GCP a few times for little side projects in the past. Knowing that they'd be prepared to shut down GCP if they're not market leader or market second within a small amount of time ensures I personally will never, _ever_ use them for enterprise work, and I'm sure I'm not alone in that.
The cloud tools do not usually have real parity or interop with other cloud tools as claimed (eg interop with boto or s3), many gcloud utilities have way worse performance and memory issues, the docs are plastered with up-selling ads and, inconsistent service account auth vs IAM-style user auth issues, and on top of all that, there is this risk Google shuts it down.
Even if they don’t shut down GCS or GKE core stuff, there are so many peripheral products like managed big data solutions, Cloud Functions, etc., that seem like _classic_ cases where Google will abruptly shut them down.
https://cloud.google.com/terms/#deprecation-policy
https://cloud.google.com/terms/deprecation
That memo was just the final nail the GCP coffin. Google wouldn't care a second about you, so why should you give them your business?
Google maintains strategic leadership in infrastructure to efficiently run Ads. The ultimate addressable market in the cloud business is as large or larger than the ads business.
Cloud also is just getting into the middle game, where commodity storage/compute services start to give way to specialized and domain specific storage/compute/analytics/intelligence. Google has unique strengths there, and a number of opportunities/advantages vis a vis Microsoft/AWS.
But rather than dominating in Cloud, Google treats it as a hobby- like TV was for Apple for about a decade. Google can perfectly well continue to treat it like a hobby. They are well ahead technically, and have plenty of cash to maintain that leadership underneath Ads.
Those resources are reporting on the framing that- for the right leaders inside Google, there is an opportunity to turn their infrastructure capabilities from a strategic skillset into a leading business. That's all.
Maybe they will, maybe not. But even if they don't, there will be no "shutting down" on any near term time frame. Even as a hobby Cloud will be big enough to be alive for a long time.
I suspect there will be an official response to the response to this pile of crap at some point.
And speaking of crap, how do you get from "risk loss of funding" to "shut down"?
[0] https://www.lastweekinaws.com/blog/google-cloud-is-probably-...
> [1] The Information: https://www.theinformation.com/articles/google-brass-set-202...
I don't see anything mentioning that Google will shut down GCP unit in 2023 - I see something saying that GCP unit will risk losing funding - presumably this is not quite the same thing. Losing funding to me would mean they will have to fund themselves on their own revenue.
Bezos himself called this a business miracle (the seven year head start, that is): https://youtu.be/f3NBQcAqyu4?t=2133
And this is against Amazon, which is known for execution. I suppose we're in for a fun couple of years, especially with the JEDI contract soap opera and all that.
looking at the history i'd say it is in MS DNA to be the successful second. In many instances they overtook the first.
However, in everything after that very first stage they were an amazing second mover: DOS, Word, Excel, etc. And now Azure.
To Microsoft's credit they maximized this advantage very well.
now we have Azure and its the same thing all over again. Azure Cloud has lots of tenancy, but these numbers I suspect are bogus. Azure runs Office 365, Exchange, Bing, and Xbox Live assets as well, and these are all "tenants." MS also bundles discounts for business critical things like desktop and server licenses into accepting Azure credits, so its likely many of these tenants are on paper only.
That being said, they've been closing the feature gap, and have shown themselves able to pick up large contracts (controversy over JEDI notwitstanding), so the market certainly seems more competitive than a couple of years ago.
As for factoring discounts etc. into cloud numbers, I confess that I don't have a good answer.
> Azure runs Office 365, Exchange, Bing, and Xbox Live assets as well
I didn't take "runs [...] assets" to mean runs the entire thing.
IIS was always a good web server. Classic ASP had unique ways of shooting yourself in the foot... but so does everything else. And in those olden days a retail box of NT Server barely cost more than a Stronghold license.
I use Azure outside the US from central Europe and the performance of their cloud services is just terrible.
Personally, I don't really like the move to the cloud at all. The whole MS palette of services from the OS to Office doesn't really convince me.
I am surprised the cloud market is actually that small. Especially since classic hosting is just included. And a growth of ~250% compared to last decade? A decade is pretty long...
Our business "jumped" into the cloud because of the reliance on MS office products. It was not done out of love.
It's not that Microsoft is stealing customers from Amazon. It's that both companies are brining in lots of new to the cloud customers, and Microsoft is just doing it faster.
But both businesses are growing at enviable rates.
Besides I’m sure that AWS and Azure are some where in the right top corner of Gartner’s magic quadrant.
The problem is Google can’t make up the market share from only pressuring “lackey” companies that it has leverage over to use their cloud, it has to have a big scale draw for otherwise unaffiliated customers, and to a lesser extent can’t sustain these initial volume & bundling discounts indefinitely.
http://www.cloudcomputing-news.net/news/2018/may/30/office-3...
Also being the discount cloud provider is not an enviable position.
I think Google has done a huge push into education, that alone could reap significant gains in 18 months.
AWS SDKs, documentation and support for Windows/.Net Framework is excellent.
I’m not making any technical arguments against Azure. I haven’t used it besides what is now “Azure Devops”.
As far as Kubernetes. It’s inconsequential when a major corporation is considering which cloud provider to use. Most existing workloads on prem are not using k8s. That goes back to knowing how to meet the customers where they are and knowing how to do a “lift and shift” as phase 1.
https://medium.com/@cfatechblog/bare-metal-k8s-clustering-at...
Edit: Way past the getting-on-board stage, apparently. It had slipped my mind that the article I linked is from 18 months ago and already described an extensive rollout.
MS has decades of building relationships with large companies that it can leverage.
If Google’s only proposition is that it’s cheaper, any AWS Solution Architect can negotiate lower prices from AWS if they want to secure a client that’s large enough.
Also GCP already does have some large enterprise clients. Some of them have had a symbiotic relationship with them, working closely to create the features they need as they scale. That kind of treatment is hard to get with AWS where most enterprises would be small fry by comparison.
There are also enterprises that don’t want to be on Amazons stuff for reasons.
All you have to do is watch the thousand plus reinvent videos to know that AWS does work with many major companies. They also have an extensive partner network.
Microsoft is breaking apart what WAS Windows Server, into discrete managed services. Identity Services, File Services, etc. If you have a need for what was formerly Windows Server provided stuff, and youre rebuilding infrastructure, you would get a lot of compatibility from Azure. Then, as you need other compute and storage resources, youre building on top of a service contract you already have, instead of spreading your data all over the clouds.
Few people get to be CTO without knowing the politics and perceptions involved with dealing with corporate America.
That could be a case for startups but it is certainly not the case for established companies.
The job of a CTO of an established company is not piloting a formula 1 car. Rather, it is to prevent reckless technical decisions from taking an ocean liner and smashing it into an iceberg.
You also need to take into account the cost of changing and tuning your code to run on a TPU.
https://dawn.cs.stanford.edu/benchmark/#imagenet-train-cost
Oh and amazingly enough, I took the exact same Python file that I built on GPU on my workstation, ran it in Google Collab with a TPU connected, and I really only had to change the configuration to make it work. There was no need to change or tune anything. I was pleasantly surprised by how well that worked.
And let me rant a bit longer, when we started this project I actually first tried to use AWS because our company already had billing set up with them. But AWS was using CUDA 10.0 for SageMaker, which has a known crash- and nan-inducing bug for 2D convolution. AWS support being non-existent as always, we couldn't make things work on AWS and then went bare metal for our GPU training with hetzner.com - which simultaneously also shaved like $30,000 monthly off our bill. AWS's EC2 GPU instances are really expensive when compared to a dedicated root server.
MS recognized moving enterprises with boring workloads to Azure was where the money was along with hybrid computing. That’s how it was able to succeed so quickly.
Besides,Microsoft has always been focused on the dark matter developers doing boring stuff. Their tooling is far better than Google’s and their IDEs that developers are already using integrate with their cloud offerings.
Also, if the only advantage that GCP has is price, that’s not defensible. It’s not like it takes a great engineering effort to change a price.
The market leader can usually afford to be a “fast follower”. That was MS’s playbook for decades.
I posted a link earlier from a podcast interview of someone who works at GCP. She said that only a few small internal projects at Google run on GCP.
This was critical because by then if you weren’t on AWS, you probably didn’t have the DevOps team to do so - MS was now willing to lend you enough consultants to do it for you.
The number of spend shouldn’t be astronomically high. Most workflows in cloud computing are commoditized low-margin. It also explains why Google wants to do high-margin AI stuff, and why AWS is happy with low-margin areas.
Microsoft would say:
“If you have a workload that still requires Windows 2000 and SQL Server 2003. We will help you migrate it to Azure as is and we will extend your support contract.”
While Google will talk about migrating your entire workload to K8s.
Part of Microsoft's implied message in heavily marketing K8s is that they know GCP is heavily K8s-based, and know GCP is heavily competing with them on price, and they are counting on some of that "dark matter conversion work" wanting to move towards the "more portable solution" in order to "maybe when the moons are right eventually shift over to GCP or something cheaper than Azure". By offering that conversion path at all, and knowing Enterprise's weakness for sunk costs development over from-scratch rewrites, they are playing a very different game than what GCP can offer.
Even if you did everything in a “cloud agnostic” way, once you have even a couple of dozen VMs, your networking infrastructure and security rules, your user permissions, your DNS entries, your site to site VPNs, your client to site VPNs, all your data in databases, security and compliance audits etc. migrating your infrastructure is going to be a months long process, involving project plans, regression tests, and still have little to show for it.
I was involved in a much smaller migration as a dev lead when all we had to do was migrate to WorkDay for HR, payroll, and time tracking. That involved tons of integration work and hiring Workday consultants. Not to mention consolidating all of the other SAAS solutions that we had to integrate when we merged with another company.
It's why Azure's K8s strategy is so fascinating: precisely because the more they make it "cloud agnostic" on paper and are good K8s/Docker citizens the greater power they have for Manager mentality lock-in to the non-cloud agnostic stuff (especially "lift and shift").
Also, a huge amount of cloud costs are based on egress, so having it on one cloud saves you resources.
All the major cloud storage providers tout geo-replication so that if an earthquake or meteor takes out a datacenter there’s still other copies.
But nothing can protect against an administrative SNAFU where you - or your cloud provider - and accidentally or intentionally deletes your account and everything’s gone in an instant (yes, there’s usually recourse and backups to recover from - but that’s hours or days of downtime).
Sometimes it isn’t even the cloud provider’s fault: see Adobe’s blameshifting onto the US executive for their dropping all their Venezuelan customers.
While Azure has the “Black Forest” unit which is legally - and technologically - firewalled off the main Azure cloud - and you can also license Azure Stack for running many (most?) of Azure’s platform on-prem or self-hosted, I haven’t seen similar offerings from AWS or Google.
To your point though, yes there is business risk of having it on one cloud.
The trick is to buy both MSFT and AMZN, and relax!
Most businesses like to keep as much of their money as they can. They can use it to pay people and stuff.
10% savings is almost never worth engineering effort. Those engineers are probably worth more than 10% by implementing features or fixing a bug for a customer.
However, even if 10% savings is $1 million, the edge odds probably still aren't in your favor. You may burn engineering time and the project fails or takes 3 times as long. Those engineers are effectively "dead" during that time and whatever contribution they could have made elsewhere is lost. For some reason, nobody ever factors that into their "savings projections".
Obviously, if I've reached the point that 10% savings is $10 million, I'm in a different bucket. But having $100 million in cloud spend probably indicates a lot more problems than AWS vs GCP.
On the other hand, some stuff is not quite there yet. The metadata server to handle instance/pod IAM (i.e. Workload Identity) is wobbly at best. To be fair, it's in beta - but then most of their offerings seem to be - which then means you can't rely on it for business-critical applications since it may wobble at the wrong time and there's no SLA.
This is just one example off the top of my head - I'm on holiday and not making massive mental effort to remember work. :)
AWS is still my preferred cloud provider. Their APIs can be daunting at times and you have to build off many lower level concepts until you get something working but when you do it tends to be rock solid - and under your control. I prefer their approach of giving you the best lowest level services to build on, and then building on top of those themselves for those who want more abstraction than the Google way of "we're building something you'll want to use - we'll tell you when it's available but it will probably be another year or two until it's GA and covered by any SLA."
This meme needs to die. It is not true and it has never been true that AWS was built out of excess Amazon infrastructure.
https://www.networkworld.com/article/2891297/the-myth-about-...
Also, genius tactic, MSFT gives you free SQL Server, Windows Server, and Exchange when you buy Azure credits. I have friends with expiring Azure credits coming up soon, waiting to see how MSFT handles. MSFT has massive cash, so prob extend for free or some other magic move.
Edit: For context, these are multi millions dollars in Azure credits, some 20M+, not being used.
I am reminded of Paul Graham's description of Yahoo's advertising business in 1999. Lots of paying customers, things seem great, you're solidly profitable. But what if most of your customers aren't profitable? What happens when the next recession comes?
Not saying cloud computing is going away, but I think its current apparent growth rates are being propped up by a lot of Softbank (and similarly unprofitable) investments, which will go away fast when the next recession comes. Until then, we don't really know how profitable cloud computing is as a business. It may not be as good as it looks right now.