Anthos – Build and manage modern hybrid applications | Hacker News Reader

Anthos – Build and manage modern hybrid applications

cloud.google.com

150 pointsroopakv63 comments

> Anthos is a subscription-based service, with the list prices starting at $10,000/month per 100 vCPU block. Enterprise prices then to be up for negotiation, though, so many customers will likely pay less.

> Google will offer a single managed service that will let you manage and deploy workloads across clouds [...] This is Google, after all, managing your applications for you on AWS and Azure

from: https://techcrunch.com/2019/04/09/googles-anthos-hybrid-clou...

lol, guess their new CEO did use the "Oracle Playbook" as he promised...

Their new CEO is an unmitigated disaster. Either he will have to change his ways or he will be leaving very very soon.

I wonder why they make things so crazy expensive or aren’t upfront about enterprise / volume pricing.

Sometimes things are negotiable, and if you list a price, it could be a turnoff for a potential (usually large) customer.

Like if you see that something costs $X, and you can only afford $Y, you might not even bother with it.

But if you talk with them they might be able to work within your budget.

But of course, when you do that, it's a clear sign you're targeting large clients/contracts, so your smaller customers won't want to bother with it.

GCP is scaling up a huge outside sales org (expensive). That means any propsect that's not an enterprise is not worth going after. Easy to weed out the medium and small businesses if you are upfront about a minimum 10k/month commitment.

I don't know – AWS has done fine with relatively transparent pricing. Telling folks you must spend $1,200 a year per vCPU for a software management layer seems ill conceived.

AWS runs a different sales model. They are much more balanced inside vs outside sales. They have inbound marketing and actively target businesses of all sizes. Inside sales is cheaper and can support smaller customers. I can tell you, AWS is not sending an experienced field rep to visit your one-man basement-built mobile app startup.

Enterprise customers have a very different idea of what crazy expensive is... if this means we have to hire one less person to run ops, this is a cost savings.

Lol. I was wondering why there was no download link.

All the pages have "Contact Sales" as the first call to action.

which is code for "This isn't for you! You can't afford this! Go away!"

"If you have to ask the price, you can't afford it."

Not to mention that it requires other licenses and hardware to work in the first place, so your total licensing cost is quite a bit higher than that:

"What's required to run Anthos? Anthos has multiple components. Among these, running GKE On-Prem requires vCenter 6.5 in order to create VMs for the GKE On-Prem cluster. Additionally, GKE On-Prem integrates with F5 BIG-IP load balancers in order to provide layer 4 load balancing."

At least AWS Outposts and Azure Stack include all the hardware and all the licenses in the box.

The Google Next Keynote (where this was just announced) is emphasizing the multicloud/cloud agnostic/on-prem capabilities, which given the emphasis on cloud lock-in nowadays is not what I expected.

It's Google's competitive advantage, and is still ultimately about lock-in.

Buy in to Kubernetes, buy in to Google Anthos, have the appearance of flexibility, but get used to GCP's tooling, quirks, documentation, etc, etc, and before long you're considering GC as the first option for all new development.

It makes perfect sense from where they are in the market.

Customers liking your service is not lock in.

...but customers being trained on “GCP's tooling, quirks, documentation” and not your competitors’ is. Also it extends beyond training, even cloud storage apis and tools are not drop-in compatible in clients and scripts.

I never understand the constant railing about "lock-in" here. My company currently runs on AWS and GCE, and maybe will be using a 3rd in the future. We can move between any provider as long as we update our tooling to work with cloud provider specifics. IE security group differences between AWS and GCE, etc.

They aren't doing anything to block us from leaving, and anyone here thinking there is going to be some kind of 100% unified API between every cloud provider is out of touch with reality, if that is what their expectation of "no vendor lock-in" is.

The real lock-in comes from using proprietary, cloud-specific services like Kinesis, Dynamo, or Spanner.

Sure, but there isn't a gun to your head. Use mysql or postgres on each provider if you're worried about that.

The current trope I see going around is that lock in centers (almost) 100% around identity management and access control, if you control that, you control your customer’s destiny. You can move a key value store anywhere on a whim.

Even just using IAM to grant permissions is going across those services is going to lock you in.

You're attempting to weaponise the notion that people are better at things they use vs things they don't use.

This can very much be a thing. Azure was originally built to look how many IT orgs deployed and managed their windows based infrastructure. It made it less scary to move into Azure, as opposed to AWS.

You can also see this in healthcare with Epic. They use absolutely bizarre names for everything to the point that it is hard for organizations using Epic to talk to organizations using other EHRs, or for employees to move between them. And since they have an absolute stranglehold on training and documentation, you basically have to go through them to be indoctrinated.

Folks have a real fear of change and learning is viewed as difficult and/or annoying. This can create considerable inertia that you can exploit.

Are you implying that leveraging that notion isn’t a business strategy?

You make it sound like lock-in is some unique strategy. That is the goal of every tech company. Enterprise, social or direct to consumer.

I think there's some nuance here that you may be discounting. There's a difference between "the appearance of flexibility" and actual flexibility. GCP's OSS friendliness and integration is legitimately a competitive advantage, but it also is being designed explicitly to avoid a need for lock-in. The cloud market is so enormous it's not necessary for Google to create explicit bias and technical stickiness in these core products.

On the other hand, where you do get potential stickiness are in the differentiated managed service offerings Google Cloud offers. These range broadly, from industry solutions & APIs to AI/ML services, to more more core data products (Spanner, Dataproc, BigQuery).

The second set of products/services don't really have any bearing on the OSS strategy and is unfair to position Google as being intentionally disingenuous in the marketing.

Anyone seen any more details on Anthos Migrate?

Google is launching the first beta of Anthos Migrate today. This service will auto-migrate VMs from on-premises or other clouds into containers in the Google Kubernetes Engine. The promise here is that this is essentially an automatic process and once the container is on Google’s platform

Edit: They demoed it on stage a few minutes after I asked... should have just waited. >.<

I'm not sure what the logic here is compared to AWS Outposts and Azure Stack. AWS Outposts and Azure stack support a very large number of Azure services and Anthos just supports GKE and Istio? So there's no first class story for basic things like queues and blob storage? How many people really use Kubernetes without needing a persistent volume claim?

I like the idea of something you can just install and run on your existing hardware, but that doesn't seem to be the goal of this either as they only support a single router and a single proprietary virtualization platform. That seems even worse than just giving you a box you can plug in like AWS and Azure.

Bigger question is what is it going to take for Google to offer anything more than this? Google has very restrictive source code policies internally, so I'd guess they'd be extremely reluctant to let the binaries for things like BigTable which run all of Google and GCP leave their network. Do they rewrite all their services or let BigTable run outside of firewall and throw an atomic clock in the box?

The same limitations apply to the other on prem services. Theres no way the real EBS control and data plane is running on prem, its just some compatible service.

Outposts has a limited feature set if you arent using vsphere, and Azure uses Microsoft's existing private cloud/on prem stack.

Enterprises are scared of one way bad decisions so if GC can convince them containerizing and moving their apps to kubernetes first is worth it over "lift and shift" this could go well. One of the few competitive advantages google has for enterprises is they can point to Amazon and Azure being forced to add managed Kubernetes services, because customers wanted that instead of a proprietary orchestrator.

Interesting. Sounds a lot like my OSS project Talos: https://github.com/talos-systems/talos. Very similar goals too.

Did they also steal your logo? Those are pretty close.

What is the story here (andrew) ?

Google has decided to imitate AWS for ridiculously non-obvious service names.

They are total legit if you speak Greek. Now engineers get to know Greek words like MDs :D

So is Google deciding to use Greek names for all their cloud services now?

Hahaha and my last startup was called sfalma (σφάλμα) which stands for error but nobody could pronounce. Unless it became a Google project. Kubernetes is pretty hard core too

In the fine Graeco-Roman "Pentium" tradition, yes.

I'm a little confused by their landing page, is this a product/packaging made for on-prem users only, or some general GKE config management that applies to all GKE customers?

Response to IBM's strategy re: the Red Hat acquisition?

No. IBM is flailing around trying to innovate it’s way out of real financial constraints that affect their bottom lines. They’ve managed to stay relevant only through acquisitions; most smart engineers have a pretty negative view of them, for good reasons: it’s not an engineering first company, it’s a sales first company, and it won’t survive for very long when most people figure out their Watson AI is a hot pile of garbage.

so this will run on AWS and Azure too.... what is it going to use to provision the cluster, kops?

Wait, isn't this what docker is theoretically capable of? (i.e. via docker-compose to manager docker containers)

Not even remotely as docker-compose is not Kubernetes, or anything resembling a container scheduler or Kubernetes Management Platform.

How long until Google cancels this with 6 months notice?

Forewarning... I think a lot of people on HN (myself included) are a bit tired of this joke now. While I get and understand the frustration/sentiment, I think it's a bit played out, and doesn't add to the discussion anymore.

It's true that it's not a funny joke, but it does make a point. Any time Google announces something cool, the first thought most engineers have is, "Do I really want to invest in this given that it could go away with minimal warning?".

It's a huge problem that Google will have to overcome. Say what you will about AWS, but they've only ever cancelled one product as far as I know (SimpleDB) and they gave people multiple years warning and a replacement product.

This case is a bit different because it's explicitly an enterprise product (w/ contracts) than a consumer product like Inbox and Reader.

Sure, but that in some sense makes it worse. So now I get locked into a contract for years on a product that maybe they'll just stop working in 1/2 way through my contract. Sure, they'll have to keep it running for me until the end of the contract, but they may not keep it running well.

Free consumer products != GCP

I understand that but it's still a reputation that follows them.

Which paid product has Google cancelled?

Google search appliance. Specifically made for the Enterprise.

That's one product. In hardware. With a 3-year deprecation phase. I can clearly see how that would be terrifying.

The terrifying part was that they stopped maintaining it years before it was cancelled and provided no replacement product.

I think GCP hasn't cancelled products either? The comparison would have to be to Amazon as a whole. And they must have a graveyard just as big as Google's.

Remember when Amazon used to run a search engine? No? They were really big into it. I even used it for a while, around 2006. Then they cancelled it, with no replacement. Remember the Fire Phone? Unsurprisingly they quickly discontinued it with no successor. (And before you object... By the standards of the "Google kills all products" meme not making a follow-up to a hardware device counts. And this wasn't the Fire Phone riding slowly into the sunset with a deprecation period.)

They had competitors for Paypal, for Yahoo Answers, Amazon used to have a Paypal competitor too. They killed it with no replacement. How about that Yahoo Answers competitor they had. Guess what? Dead. With no replacement. Amapedia, their "wikipedia for products sold on Amazon" is of course dead. Because it was an Amazon product, and that's what Amazon do. They make products, and then they kill them.

But somehow you're willing to give Amazon a free pass, while thinking that the worn-out joke about Google is a "valid point".

Not everybody is on the same page w.r.t. "dead memes," for example I had not considered Google's history of service cancellation when I saw this page.

This outdated joke doesn't even apply to Google Cloud anyways. In its 10 years of existence, they haven't killed a single service.

People comparing the deprecation of an RSS reader to that of an enterprise-tier service clearly don't understand anything about tech.

It's not just the specter of cancellation that haunts some people's perception of Google; remember the big price hike of GAE? The point is that Google is very unpredictable and in the "Enterprise" world that is a huge no-no.

When will it be discontinued?

Probably when Reader is brought back.

And it will be branded as anagnostes

Reply on news.ycombinator.com