Performance on single-bucket multicalls

our company is going to use storj for storing data, splitting bucket for multiple parts with different access, so different people could use it in parallel without having access to each other’s data.
what would would be better approach in terms of performance: split by separate prefix or passphrase?

Hi @anna.ivaniuk

Welcome to the Forum!

This documentation may be what you are looking for


I do not see anything in these documents specifically mentioning performance. As a node operator I am not familiar with satellite code, but I can talk about raw storage itself. Currently storage nodes are selected independently from prefix or passphrase for a given bucket, so the raw bandwidth or latency when using the uplink protocol would be the same regardless of the choice of prefix vs. passphrase. What’s more, storage nodes do not know either, so they cannot make any decisions based on them.

What may be different is metadata operations, for which satellites are responsible. Someone from Storj Inc. would probably have to answer this part of question though.

1 Like

Yes, using a different passphrases or/and prefixes shouldn’t affect neither latency nor speed, both are depends on location, your bandwidth and hardware power (CPU/RAM), it also depends on used protocol - if you have a wide bandwidth, the native protocol can be better, if it’s limited or you do not have resources on the device where you plan to use uplink, then you can use S3 protocol.

1 Like