We have a educational system (federated microservices that are running mostly on AWS EC2, RDS MySQL and a message queue) that is already distributed across regions, primarily for latency hiding, though it has been useful for failovers. We're now considering adding HA within each region, by adding active-passive failovers in using multiple AZs in each region. However, we received an attractive offer from another large IaaS provider to host the failover on their infrastructure at a very competitive price point.

What could be the downside of having a deployment that uses multiple IaaS service providers? There isn't much literature out there from real-world deployments.

  • Latency, having to learn the ins-and-outs of two different large systems, maintaining two sets of permissions and API calls, etc. There are positives too, of course. – ceejayoz Jul 9 '16 at 15:59
  • 2
    I can see benefits as you won't have all your eggs in one basket and as you can already see, you can benefit from the competition between different cloud providers. Doing so paves the way to also move away from any supplier which might be invaluable in the future. The risk is of course that you increase complexity because you need to deal with idiosyncrasies inherent in each specific platform. For real world architectures the High Scalability blog is interesting: i.e. this – HBruijn Jul 9 '16 at 15:59
  • Architecturally, the way we have sketched it out, it should be possible to avoid most "cross-cloud" traffic, which should prevent most latency increases. The problem is the DB... :-( – Jedi Jul 9 '16 at 16:25
  • @HBruijn yes that's very similar to what we're looking at. Keeping the data strongly consistent is probably something that we are going to give up. – Jedi Jul 11 '16 at 16:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.