Hey folks, I had some question around hybrid cloud...
# platform-blueprints
Hey folks, I had some question around hybrid cloud workloads. I’ve only got experience either managing on-prem infrastructure or for the last decade or so, cloud-only. Has anyone got experience providing platform tooling around multi-vendor or hybrid cloud environments? Some insights I’m interested in are: 1. What are some of the challenges you have VS cloud only? 2. What tooling options are better suited to a hybrid model? 3. What are some of the benefits you have VS cloud only(beyond cost)? 4. How have you found managing these? Is going Cloud only worth all the extra costs or actually on-prem also was surprisingly easy? Any and all answers would be super helpful. Looking to get more into this space so any resources or recommendations would be awesome too. thankyou 1
Hmm… what exactly is a hybrid cloud workload for you? Do you really want to e.g- distribute parts of an application over multiple clouds, or do you want to run different cloud estates? The first is pretty hard, the second is pretty trivial. Problems usually arise when you either want tooling that allows you to address both your estates in an uniform way or when you want to be able to move applications between clouds without much effort - hint… they need to be desined and built for that in the first place. Going hybrid - cloud + on-prem can be pretty trivial if you buy into the cloud provider model of connecting the two estates. Having baseline network connectivity makes most problems disappear. If you go without that it gets hard to get things right in a secure and reliable way. I don’t want to pitch here, but you could take a look at the reference architectures we provide - you can have many “resouce planes” as you like, turning that into a multi- or hybrid scenario immediately. Feel free to exchange any logos in there for ones that fit your world better. You can get inspiration from the community tooling landscape.
The topic is non-trivial - happy to spend some time discussing over a virtual coffee if you want @Shane Dowling
Thanks for the insights @Clemens Jütte! So it’s not something my company is looking into, it’s more I get a sense that with companies becoming more cost-sensitive, the idea that Cloud providers are where you should run everything is starting to get challenged. Which may change the platform engineering landscape. So what I’m thinking more about is hybrid cloud where you’re distributing workloads between on-prem and the cloud. Practically speaking I don’t think I can see much value in blindly distributing across both but more tactically using the Cloud for more flexible workload deployments and on-prem for more static workloads. I was curious if they need entirely different technology choices and if/how people build consistent abstractions over them.
it really comes down to a selection / choice on a case by case basis
if you are thinking a purist experience of consistency of application runtimes and workloads across estate types, you are likely going to land on non-SaaS offerings or their private data center versions for each component, e.g. Elastic Stack vs Elastic Cloud
but yes, it's a very broad and deep topic that we likely can't cover easily in a slack thread since it also touches on things like culture, skills, etc.