Hello, question about parallel usage. Is there an...
# gooddata-cloud
s
Hello, question about parallel usage. Is there any way or any concept where users have more than one set of compute resources tied to their data source?
We're snowflake, but stating that to ask more question. I ask this for two main reasons 1. We are running into what appears to be performance issues with the vast number of customers using our workspaces at the same time. Therefore having the ability to say use more than 1 connection or warehouse might help 2. Has anyone attempted to shard their customers or workspaces to even out their load of customers a. ie we have 3000 workspaces we create 3 shards of accounts with each workspace in parallel environment configs.
r
Steve - we're discussing this internally and will respond in the next day or two. Cloud DW performance acceleration and cost reduction are the design goals of GoodData FlexQuery
s
Thank you for the response. Very interested on how that could be used to proceed.
🙌 1
Definitely wouldn't want to have to build something like multiple parallel environments so 100% excited to learn about it.
m
Hi Steve, currently a single GoodData datasource does not allow to add multiple Snowflake warehouses, but there are several options how to achieve scalability: • one purely on the Snowflake side might be to utilize the Multi-cluster warehouses and scaling policy that allows to scale the compute either manually or automatically based on the load while having one warehouse. • or you can use the newly released option in GoodData - datasource per client where you can create additional datasources (i.e. with different Snowflake warehouses) and assign them to some of your client workspaces to override the datasource from the parent workspace. This way you can have full control over which warehouse is used for which client workspace and for example isolate your biggest/most important workspaces. We are also planning further performance improvements to our FlexQuery engine in the future to bring better performance (i.e. pre-fetching of some data to speed up the calculation etc.)