6 ms·

I do wonder if a cloud bubble is actually popping. For small orgs you have Heroku, etc. Having a dumb place to run code or turn on a service is really valuable

I do wonder if a cloud bubble is actually popping. For small orgs you have Heroku, etc. Having a dumb place to run code or turn on a service is really valuable. Especially when it comes with an SLA and support. Above a certain level you probably adopt k8s for different components, and just want to treat your cloud as a commodity. Then you realize, like base camp has, any kind of compute - even static servers you own - also does the trick and can also run k8s.

scarface74 4y ago
You realize Heroku runs on AWS right? And no, most organizations are not using K8s and never will. Docker is not a magic bullet.
- mlyle 4y ago
  > For small orgs you have Heroku, etc. Having a dumb place to run code or turn on a service is really valuable. Especially when it comes with an SLA and support. Yes, he does. He's saying that when you grow to a point, you might want to seek just dumb commodity compute -- even including servers you own.
- Spivak 4y ago
  Docker, well the OCI and their specs, isn’t magic but it is absolutely the magna carta of application packaging, distribution, and deployment. Docker commoditized “platforms” and k8s is on its way to do the same to “cloud.” And it’s a good thing, as someone who is very on-prem oriented. I now have a set of APIs and specs that let me present “here’s your full self service cloud devs, enjoy.” And all the tooling works.
  - scarface74 4y ago
    You don’t work with the large enterprises that I work with everyday. Let’s not pretend that K8s doesn’t bring its own set of complications. Not to mention enterprises that have 500+ servers, some running Windows, some running COTS, etc. Much of the time it is easier just to spend up a VM and install whatever you need.
    - Spivak 4y ago
      Ahh government work. Can always tell because y’all don’t call them appliances. But I mean idk, I kinda do. The largest place I ops’d for had two 40 rack floors with some misc stuff here and there. Don’t remember at all how many of individual servers that meant once you accounted for KVMs, SANs, UPS, and networking but it was a lot. I mean yeah we had a huge Hyper-V cluster for the Windows IT folks, bunch of random appliances bought over the years, but then for our software stack we ran a big-ole RHV cluster (look, we had a sweetheart deal from RH gave us everything) spent a lot of time getting k8s hooked up to everything (open stack we decided wasn’t worth the squeeze) and devs were v happy, no more tickets! All the backing infrastructure was vm based but we managed it and the devs got to care no percent.
      - scarface74 4y ago
        > Ahh government work. Can always tell because y’all do Uhh I work in the Professional Services department of a little cloud provider based out of Seattle you might have heard of… Sure I’ll work on a project porting to EKS (managed Kubernetes) if they are already familiar with it. But rarely will I suggest introducing the complexity of it unnecessarily. I’ve worked with governments, colleges, F500s, startups, and everything in between. I mostly focus on green field app dev. But I know my way around “DevOps” [sic]
        Spivak 4y ago
        Ahh government work! Look I’m just poking fun at the fact that people who call them COTS are either government or sell to government, it’s not a pejorative. Not doubting your expertise at all, I’ve worked at shops that are entirely vm based, I just think that when it comes to making dev teams happy in an on-prem world giving them a k8s cluster has so far never not been a good decision. I can hide so many of the ugly realities of running a production service.
        scarface74 4y ago
        You have to talk the talk to show them that you’re one of them. I can go from a decent midwestern neutral accent to my natural southern drawl depending on which state I’m working with too :) And besides, even when I worked in the real world as your bog standard Senior enterprise dev, I got to the point where I wouldn’t work at any company that had on prem architecture. My main reason for going to the cloud was not that I didn’t want to deal with administering servers, I didn’t want to deal with server administrators.
- bamboozled 4y ago
  Kubernetes and Dockers aren’t coupled anymore…
  - NathanOsullivan 4y ago
    Like it or not, Docker is the Kleenex of containers.
itisit 4y ago
Popping? Not for AWS. Still posting robust double digit growth YOY. Most of the cloud abstraction services out there rely heavily upon AWS.
nytesky 4y ago
It does make me think about networking companies and dot.com. Fiber and networking gear considered themselves “real” businesses versus those “web” companies that imploded. Of course, all that IPO money had been buying their hardware and there was a echo crash and a decade of dark fiber. AWS may have a significant customer base of VC funded startups which may be scaling back their spending rapidly; coupled with larger customers migrating to on premise it could eat away at both ends of their market share.
- bushbaba 4y ago
  The issue is that AWS has double exposure. Startups and B2B company's selling to startups (e.g. datadog). However, AWS will be fine. Their growth rate is what 30% YoY. I highly doubt the reduction in spend to comprise 30%+ of their revenue base. AWS Growth rate over time. Note while growth rate declines, the amount of revenue added each quarter increased. (Percentages are easier to grow when you have a small base) https://www.statista.com/statistics/422273/yoy-quarterly-growth-aws-revenues/ https://www.statista.com/statistics/422273/yoy-quarterly-gro...
tinyhouse 4y ago
Not a bubble in terms of cloud adoption but a lot of customers reduced their cloud bill. What happened was that many companies didn't invest time and effort in being more efficient in how they use the cloud and suddenly when money is tight everyone realizes they can save a lot of money by being more responsible. For example, don't use the most expensive GPU instance when training your ML model. Kill the instance once you don't need it. Train smaller models. Etc.
darkcha0s 4y ago
When I read these type of comments, I wonder if OP is living in some kind of bubble. Large corporations spent billions over the last years migrating to the cloud. You can't simply 'undo' that. All fine and dandy for small orgs, but they aren't the ones AWS makes most its money with.
- Spooky23 4y ago
  Cloud requires growth. All of the massive capital investments require more and more growth to be sustainable. Even a small shift in the demand curve is very expensive for them. The response is to cut costs and raise prices.
- softwaredoug 4y ago
  I would guess companies spending billions migrating to clouds has probably peaked, and that would also be part of an argument for cloud bubble bursting.
  - darkcha0s 4y ago
    So you foresee the collapse of swaths of large corporations? Or what happens when this bubble bursts
unethical_ban 4y ago
Last employer is banking. They have a multi year investment in moving various workloads, from VM compute to lambda to container, into multi region, multi cloud environments. Dozens of dev, sec, and ops people working together to make provisioning seamless and safe, code deploys done through CI. When you make that kind of decision, going back or altering course is a similar multi year decision. Going cloud native at a cost of tens of millions in billing and labor specialty means you have lost or reprovisioned other specialties, haven’t renewed on premise equipment, and so on. These companies are like supertankers, not racing yachts. Take a long time to alter course.

scarface74 4y ago

You realize Heroku runs on AWS right? And no, most organizations are not using K8s and never will. Docker is not a magic bullet.

mlyle 4y ago
> For small orgs you have Heroku, etc. Having a dumb place to run code or turn on a service is really valuable. Especially when it comes with an SLA and support. Yes, he does. He's saying that when you grow to a point, you might want to seek just dumb commodity compute -- even including servers you own.
Spivak 4y ago
Docker, well the OCI and their specs, isn’t magic but it is absolutely the magna carta of application packaging, distribution, and deployment. Docker commoditized “platforms” and k8s is on its way to do the same to “cloud.” And it’s a good thing, as someone who is very on-prem oriented. I now have a set of APIs and specs that let me present “here’s your full self service cloud devs, enjoy.” And all the tooling works.
- scarface74 4y ago
  You don’t work with the large enterprises that I work with everyday. Let’s not pretend that K8s doesn’t bring its own set of complications. Not to mention enterprises that have 500+ servers, some running Windows, some running COTS, etc. Much of the time it is easier just to spend up a VM and install whatever you need.
  - Spivak 4y ago
    Ahh government work. Can always tell because y’all don’t call them appliances. But I mean idk, I kinda do. The largest place I ops’d for had two 40 rack floors with some misc stuff here and there. Don’t remember at all how many of individual servers that meant once you accounted for KVMs, SANs, UPS, and networking but it was a lot. I mean yeah we had a huge Hyper-V cluster for the Windows IT folks, bunch of random appliances bought over the years, but then for our software stack we ran a big-ole RHV cluster (look, we had a sweetheart deal from RH gave us everything) spent a lot of time getting k8s hooked up to everything (open stack we decided wasn’t worth the squeeze) and devs were v happy, no more tickets! All the backing infrastructure was vm based but we managed it and the devs got to care no percent.
    - scarface74 4y ago
      > Ahh government work. Can always tell because y’all do Uhh I work in the Professional Services department of a little cloud provider based out of Seattle you might have heard of… Sure I’ll work on a project porting to EKS (managed Kubernetes) if they are already familiar with it. But rarely will I suggest introducing the complexity of it unnecessarily. I’ve worked with governments, colleges, F500s, startups, and everything in between. I mostly focus on green field app dev. But I know my way around “DevOps” [sic]
      - Spivak 4y ago
        Ahh government work! Look I’m just poking fun at the fact that people who call them COTS are either government or sell to government, it’s not a pejorative. Not doubting your expertise at all, I’ve worked at shops that are entirely vm based, I just think that when it comes to making dev teams happy in an on-prem world giving them a k8s cluster has so far never not been a good decision. I can hide so many of the ugly realities of running a production service.
        scarface74 4y ago
        You have to talk the talk to show them that you’re one of them. I can go from a decent midwestern neutral accent to my natural southern drawl depending on which state I’m working with too :) And besides, even when I worked in the real world as your bog standard Senior enterprise dev, I got to the point where I wouldn’t work at any company that had on prem architecture. My main reason for going to the cloud was not that I didn’t want to deal with administering servers, I didn’t want to deal with server administrators.
bamboozled 4y ago
Kubernetes and Dockers aren’t coupled anymore…
- NathanOsullivan 4y ago
  Like it or not, Docker is the Kleenex of containers.

itisit 4y ago

Popping? Not for AWS. Still posting robust double digit growth YOY. Most of the cloud abstraction services out there rely heavily upon AWS.

nytesky 4y ago

It does make me think about networking companies and dot.com. Fiber and networking gear considered themselves “real” businesses versus those “web” companies that imploded. Of course, all that IPO money had been buying their hardware and there was a echo crash and a decade of dark fiber. AWS may have a significant customer base of VC funded startups which may be scaling back their spending rapidly; coupled with larger customers migrating to on premise it could eat away at both ends of their market share.

bushbaba 4y ago
The issue is that AWS has double exposure. Startups and B2B company's selling to startups (e.g. datadog). However, AWS will be fine. Their growth rate is what 30% YoY. I highly doubt the reduction in spend to comprise 30%+ of their revenue base. AWS Growth rate over time. Note while growth rate declines, the amount of revenue added each quarter increased. (Percentages are easier to grow when you have a small base) https://www.statista.com/statistics/422273/yoy-quarterly-growth-aws-revenues/ https://www.statista.com/statistics/422273/yoy-quarterly-gro...

tinyhouse 4y ago

Not a bubble in terms of cloud adoption but a lot of customers reduced their cloud bill. What happened was that many companies didn't invest time and effort in being more efficient in how they use the cloud and suddenly when money is tight everyone realizes they can save a lot of money by being more responsible. For example, don't use the most expensive GPU instance when training your ML model. Kill the instance once you don't need it. Train smaller models. Etc.

darkcha0s 4y ago

When I read these type of comments, I wonder if OP is living in some kind of bubble. Large corporations spent billions over the last years migrating to the cloud. You can't simply 'undo' that. All fine and dandy for small orgs, but they aren't the ones AWS makes most its money with.

Spooky23 4y ago
Cloud requires growth. All of the massive capital investments require more and more growth to be sustainable. Even a small shift in the demand curve is very expensive for them. The response is to cut costs and raise prices.
softwaredoug 4y ago
I would guess companies spending billions migrating to clouds has probably peaked, and that would also be part of an argument for cloud bubble bursting.
- darkcha0s 4y ago
  So you foresee the collapse of swaths of large corporations? Or what happens when this bubble bursts

unethical_ban 4y ago

Last employer is banking. They have a multi year investment in moving various workloads, from VM compute to lambda to container, into multi region, multi cloud environments. Dozens of dev, sec, and ops people working together to make provisioning seamless and safe, code deploys done through CI. When you make that kind of decision, going back or altering course is a similar multi year decision. Going cloud native at a cost of tens of millions in billing and labor specialty means you have lost or reprovisioned other specialties, haven’t renewed on premise equipment, and so on. These companies are like supertankers, not racing yachts. Take a long time to alter course.