Does aws or azure worry about having no direct competitive product against spanner?
On a related note, can anyone on say how far they have actually scaled up spanner (as in, does it actually live up to the hype)?
Cosmos is GA and is used in “real life” by Fortune 500 companies. AWS azure and google all have good stuff. No one is eating the others lunch.
Fair enough - but if Cosmo goes down, Microsft will be fine.if Spanner goes down, Google ends as a company. Just different levels of importance and investment.
Well, we use CosmosDB all over azure now. especially now that table storage is getting merged with CosmosDB (whatever that means)
If you think about it, AWS serverless aurora couls provide sql at potentially infinite scale in the future
Spanner is very expensive
It's cause it's a super hard problem - you only need it for geo distributed eventually consistent workloads.
Spanner is actually pretty cheap for DBaaS. FaunaDB is the closest DBaaS alternative to Spanner, and is currently 10-100x more expensive than Spanner.
The way I see it is that there are very few customers that would need spanner. You would need a very high write workload and be globally distributed (AND instead of OR because the two problems combined are the difficult part, either on its own would be something that AWS or Azure could tackle quickly). The problem is that you have to rewrite your code base for spanner and you essentially lock yourself into GCP. This means that as a company that is big enough to have those requirements, you’re betting your company on googles cloud success. To top that off GCP needs to offer all of the other amenities (services) that you need to run your business. And then you have to rely on them providing good support, something that has caused huge customers to leave them. So I wouldn’t see spanner as something to worry about. It’s really impressive technology but I don’t see it as something that is going to cause customers to pick GCP over Azure or AWS.
Can you name some huge customers who have left? Because we've added a ton of AWS customers, but I've RARELY heard of the reverse. Further, there's nothing that locks you into GCP. It's just an API - you can easily add Spanner into your infra where you need geo distributed eventually consistent data - I'd NEVER use it when you don't care that much, just use Aurora, CloudSQL, etc. Finally, it's super unlikely that anyone could spin up this tech quickly. To do it right, you literally need atomic clocks, hardware root of trust, private globally routable networks with well defined latency boundaries, and more - check out the paper! https://research.google.com/archive/spanner.html Spanner supports high write workloads but that's not why you use it - it's for workloads you need a strong guarantee around how long it takes to be eventually consistent. That's the hard part - if you have a globally distributed set of readers and writers, and need to make sure they read or write with consistency, that's why you use it. Payment processing, user profiles, gaming, etc.
Like I said both combined is difficult, either on its own is much simpler. I could tell you multiple ways of accomplishing each on their own. The lock in comes from the fact that it is a google only protocol. You can’t really say you’re going to use spanner for your database while using EC2 for your compute. All of the cloud providers have this, it’s not a dig on GCP. Customers who don’t want to be locked into a single vendor must use the lowest common denominator of them all. Spanner is one of those things that isn’t the LCD which is great. However I don’t see it as something that any customer was waiting for before switching to GCP (though I’m sure there was research done to prove me wrong). No I won’t publicly mention which customers but the really big ones are here on blind with us.
I was promised SQLAlchemy support. When that happens, RIP Oracle and other RDBMS.
Also, all of Amazon The Bookstore runs on either DynamoDB proper or similar.
None of it did when I left in 2015, but some things might have changed since then.
Do you remember project rolling Stone? All tier-1 services needed to get off RDBMS and onto Sable/DDB. Both datastores added a ton of features to support all of the usecases.
AWS Aurora now supports multi-write master in beta which may be the answer to spanner's write scale-out
There's a lot of great engineering happening, so we all win. There are differences in the products but they each simplify / enable solving big problems. Competition is good for the industry and we're just getting started.
Azure has cosmodb in beta, I don’t think Aws has anything. Spinner is core to Google’s infrastructure, even if they both launch GA products, it’s unlikely they’ll have the same size staff or importance for their companies.
Cosmodb is not relational so not the same thing at all
Aws has global tables