I’m currently using the free version of Pinecone for a PoC, and we’re doing a cost analysis of which vector database service to use for the production-ready version of our service.
The Pinecone is pricing is indicated on a per hour basis - and (if I’m using AWS) - I assume it’ll be running on EC2 instances.
Generally this will be an internal-only application, (at least in the first stages) so I’m wondering if it’s possible to set up some schedule to put instances to sleep during times that they’re not needed.
I don’t believe this is directly possible via the API (there is no “sleep” or “pause” type function). But if you really needed this you could probably cobble something together using the Collections feature. Collections are like a backup of index data, so this would end up looking a bit like a backup&restore flow.
Shut down flow:
Back up data from Index → Collection
Delete index
Startup flow:
Create index
Restore Collection → Index
Some things you’d want to consider:
How much data you have (this would determine how long the shut down or startup flows take)
How much time you have available for the startup process to “wake” the index – and do you know sufficiently far enough in advance that the index needs to be woken up before it needs to start serving requests