4 ms·
This seems to get back to an interesting discussion I had with someone early in the OpenStack efforts: as a developer OpenStack doesn't directly interest me bec
by WorldMaker 10y ago
This seems to get back to an interesting discussion I had with someone early in the OpenStack efforts: as a developer OpenStack doesn't directly interest me because I don't care about infrastructure. Where OpenStack had a possibility to win was to provide options for infrastructure agnosticism: if I can build an app that runs "unmodified" on any OpenStack-based infrastructure, that has a possibility to save me potential time and money from having to port apps to/from/between AWS, Azure, Google Cloud, et al (assuming of course that it enough clouds actually adopt).
From that perspective, container solutions are delivering a better developer proposition than OpenStack has yet managed. There are ways now to build container clusters that you can ship in parallel to AWS and Azure with very little code difference.
In that earlier discussion I was skeptical of OpenStack precisely because of its focus on infrastructure first. Without the buy in of being a clone for a specific cloud structure (AWS compatibility over anything else, for instance) or the backing of traditional datacenter/server vendors (IBM who eventually started into BlueMix; Microsoft whose "on premises Azure" is now firing on most cylinders but was announced as a plan early in OpenStack's history), OpenStack didn't seem to have an obvious niche in the infrastructure world. The closest to a niche it might have had in its early life was the promise of application portability between clouds and that never quite seemed to be delivered.
I can tell it frustrates infrastructure folks to hear that containers have been eating OpenStack's lunch, but that is the very real case from the developer perspective. As a developer today, I go for containers and OpenStack is no longer relevant on my radar. Sure I can run containers on OpenStack, but containers abstract away more of the infrastructure and I have less and less care what cloud(s) is underneath the container cluster. When I asked OpenStack people what OpenStack might deliver to me that vision of application portability was tantalizing but never seemed quite finished; container technologies have actually delivered that.
- jacques_chester 10y ago> if I can build an app that runs "unmodified" on any OpenStack-based infrastructure, that has a possibility to save me potential time and money from having to port apps to/from/between AWS, Azure, Google Cloud, et al (assuming of course that it enough clouds actually adopt). OpenStack is at the same layer as AWS, GCP or Azure. It's not an abstraction over them -- that's a PaaS. I'm not sure OpenStack was ever seriously intended to work at that level. I work on a platform, Cloud Foundry, which provides the kind of portability you're talking about. Red Hat have OpenShift, which is another such platform. Disclosure: I work on Cloud Foundry on behalf of Pivotal.
- WorldMaker 10y ago> I'm not sure OpenStack was ever seriously intended to work at that level. There was a hint at some possibility that it might, especially in some of the hype circles early on. It is of course unfair to blame OpenStack for never actually fulfilling that dream, when that was not its intention per se, but that is part of why today people continue to compare what OpenStack promised versus what containers delivered.
- derefr 10y agoIt doesn't have to be an abstraction; it could have been an API standard for IaaS providers, with OpenStack itself serving as a reference implementation of said API. No higher-level concepts like containers; just saying things like "the VM lifecycle controllers in AWS, Google Cloud, Azure, DO, Linode, Rackspace, some random company's private cloud, etc., should all 'speak Nova', so that any Nova client can control VM lifecycles on any of them." Such providers could expose extra features with their own proprietary APIs (think e.g. the features exposed to the Gmail users through the web-app vs. through IMAP) but the standard protocol would at least be there for everything. Not that asking AWS+GCP+Azure to commoditize themselves in such a way makes much economic sense, but that's the proposition.
- jacques_chester 10y agoIt'd be nice, but I don't think they will. Cloud Foundry gets around this by using BOSH for deployment; it essentially creates the kind of layer you're describing.
- yeukhon 10y ago> if I can build an app that runs "unmodified" on any OpenStack-based infrastructure, that has a possibility to save me potential time and money from having to port apps to/from/between AWS, Azure, Google Cloud, et al (assuming of course that it enough clouds actually adopt). > There are ways now to build container clusters that you can ship in parallel to AWS and Azure with very little code difference. OpenStack is an IaaS, so just like AWS. Focus on the context. Do you want your own private cloud? Yes or not? If no, then this discussion can end, because AWS and Azure run on their proprietary IaaS code. As a customer, you request resources from the IaaS layer, and you build your server/platform from that point. So arguing that container clusters can ship to other clouds with very little change (almost likely writing the APIs to create a container) is unfair in the context of why one would choose container over OpenStack. The purpose is different. If your answer is yes I am building a private cloud, how are you going to do that with Docker alone? Can you build a software-defined network with Docker? Absolutely not with Docker since Docker is a host-based deployment solution. What you are looking for is ability to create OpenStack the same way CloudFoundry / Kubernetes are created. You write up a manifest, and the necessary databases and services are deployed to some EC2 machines. In the case of CloudFoundry, you write up a manifest file, describes number of instances, types, credentials, what not, then call Bosh to create Cloud Foundry (will create a pool of app machines, router servers, UAA, etcd etc). Machines are created based on a stemcell, basically an image. You want to create your IaaS based on images. You want to be able to script up a manifest and deploy your IaaS. You want a lift and drop IaaS infrastructure. Container can do that, but it cannot be done simply with Docker. You need that infrastructure abstraction layer to cover up. That's probably what Docker Enterprise Edition might do, but I have not really dig into it yet.
- WorldMaker 10y agoYes, it is an unfair comparison. I was attempting to explain in my comment why I consider it a valid unfair comparison, because it is a matter of perspective. As a software developer, do I ever want a private cloud? No. Does my employer? Maybe. Is it my job to tell them how to invest their infrastructure dollars? Quite possibly no, because software development and infrastructure are typically held at arms length. But even when they are not in a "proper" DevOps shop, the ballgame of which cloud is then subservient to developer convenience and how easy it is to deploy software to a cloud and how productive developers are writing software for that cloud. So yes, the purpose of OpenStack and Container technologies are very different and I appreciate that technically. In terms of real world value to me as a software developer, however, I have platform problems not infrastructure problems. I don't care what the infrastructure is under the service so long as it provides a stable, reliable platform for me to build upon. Containers abstract that for me in a way that solves real platform problems that OpenStack was only ever relevant to me in so far as its ability to once hint at a possible solution to. That's not fair and that was expecting too much from OpenStack at the time, but that's life.