Distributed Data Show Episode 56: Multi-Cloud: The What & Why

Today we’re tackling what confusion my exist around the various *-clouds. We have multi-cloud, hybrid-cloud, private-cloud, and so many more. After we narrow down the multi-cloud description to it’s key points the conversation begins around why we’d want something multi-cloud to begin with! It’s the beginning of a series of conversations we’re going to have around the multi-cloud space and all of the complexities, advantages, lock-in, and related topics around this new foray into distributed systems technology.

ABOUT DATASTAX ENTERPRISE 6
DataStax powers the Right-Now Enterprise with the always-on, distributed cloud database built on Apache Cassandra™ and designed for hybrid cloud. DataStax Enterprise 6 (DSE 6) includes industry-leading performance, self-driving operational simplicity, and robust analytics.

CONNECT WITH DATASTAX

ABOUT DATASTAX ACADEMY

[rjhintz]

Interesting topic. Some comments:
-private clouds, do they still exist? Presumably any prod OpenStack systems would qualify as private clouds. I understand these are especially popular in China, though somewhat in decline elsewhere.

-around 2:22, running a service, a DB across >1 provider? I'm not sure how a DB would run across multiple providers in prod, even something like Cassandra. Latencies and jitter would presumably be too much. Ok, how about a stateless service offered in >1 provider, would you want to design a production service with no latency or jitter expectation, much less guarantee?

-around 2:45, "best of all worlds." There's an argument that it's actually the worst of all worlds. Besides networking issues, you'd need procurement, staff expertise, versioning harmony across >1 provider when it's hard enough to do just one. And that's just for the most basic, lowest common denominator services. APIs are different, bug fix and feature cadences differ, added value services differ. How would one even put the pieces together and who would have the skill set to do it, maintain it, diagnose it?

-around 3:35 move workloads among clouds. This sounds a bit like the old argument for cloud as a bursting target for overstressed onsite workloads. I'm not aware that bursting ever worked for transactional workloads, but would be interested in any actual experience. Similarly, moving workloads among clouds sounds really tricky, as a practical matter. Any thoughts on how this might work?

-5:15 typical architecture, copy A in one, copy B in another. How does data consistency work in this scenario?

-5:45 If one workload uses AWS Lambda presumably against data on AWS, how can workload data also be on Azure? Two companies were mentioned, so different workloads? I didn't understand this. I'm sure I'm mischaracterizing what was said/intended.

I appreciate the discussion and hope these comments aren't the usual YouTube dross. ;)