I would say that in general about proprietary databases (not just those from Amazon!) but point made.
I’m very curious about your negative experience with DynamoDB, though. My work is talking about a long-term shift toward non-SQL—not necessarily Dynamo, but it’s a name that’s come up.
Our pain is a bit less now with a 50% cost drop on the service that AWS pushed last month but our engineers created a write heavy workload which wreaked havoc on our bill. So it’s serverless to a point but even then in certain cases either the scaling up won’t catch up on time or the billing will come to kill you, so at the end of the day it’s a lie anyway (just like all serverless is).
I’m generally not really sure what the point of Dynamo is and I feel like if you need some kind of document/no-SQL database you should probably look at Scylla which I’ve seen being used with great success.
Scylla requires a license for clusters with 5+ nodes and you need to pay for the hardware, likely scaled to max RPS + some buffer. It might be cheaper at some scale, but for many, that’s a hefty start price.
I can only guess that they are an enterprise customer of Scylla, much like Discord is [1].
Since all pricing on the ScyllaDB website is “contact us for a quote”, I suspect they are paying a hefty price [2] for high quality software
Congrats to the team - this looks very interesting. Interesting to see that they got rid of MVCC. Unfortunately the Postgres compatibility is quite limited at the moment - no views, temporary tables, foreign keys, or exclusion constraints. Maybe in due time..
As a person digging us out of a DynamoDB shaped hole, I’d just caution people to be very careful with Amazon proprietary databases.
I would say that in general about proprietary databases (not just those from Amazon!) but point made.
I’m very curious about your negative experience with DynamoDB, though. My work is talking about a long-term shift toward non-SQL—not necessarily Dynamo, but it’s a name that’s come up.
Our pain is a bit less now with a 50% cost drop on the service that AWS pushed last month but our engineers created a write heavy workload which wreaked havoc on our bill. So it’s serverless to a point but even then in certain cases either the scaling up won’t catch up on time or the billing will come to kill you, so at the end of the day it’s a lie anyway (just like all serverless is).
I’m generally not really sure what the point of Dynamo is and I feel like if you need some kind of document/no-SQL database you should probably look at Scylla which I’ve seen being used with great success.
Scylla requires a license for clusters with 5+ nodes and you need to pay for the hardware, likely scaled to max RPS + some buffer. It might be cheaper at some scale, but for many, that’s a hefty start price.
Ah I guess open source really does not exist anymore.
Do you think Bluesky is paying top dollar for their Scylla deployment?
I can only guess that they are an enterprise customer of Scylla, much like Discord is [1]. Since all pricing on the ScyllaDB website is “contact us for a quote”, I suspect they are paying a hefty price [2] for high quality software
[1] https://discord.com/blog/how-discord-stores-trillions-of-messages
[2] https://www.scylladb.com/product/scylla-cloud/get-pricing/
Congrats to the team - this looks very interesting. Interesting to see that they got rid of MVCC. Unfortunately the Postgres compatibility is quite limited at the moment - no views, temporary tables, foreign keys, or exclusion constraints. Maybe in due time..
I mean you shouldn’t be using most of those anyway if you want a truly distributed database.
Agreed for FKs and exclusion constraints for sure. At least for now.