Do not sign every request for streaming RPC with safe endpoints #942
Labels
No labels
P0
P1
P2
P3
badger
frostfs-adm
frostfs-cli
frostfs-ir
frostfs-lens
frostfs-node
good first issue
triage
Infrastructure
blocked
bug
config
discussion
documentation
duplicate
enhancement
go
help wanted
internal
invalid
kludge
observability
perfomance
question
refactoring
wontfix
No project
No assignees
2 participants
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference: TrueCloudLab/frostfs-node#942
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
If we connect to some endpoint with TLS enabled, there is no need to authenticate every single message.
This works for both client- and server-side streams.
For example, in
tree.GetChildren
orobject.Search
the amount of responses scales linearly with the number of objects in container. Signing and verification introduce latency on each operations in stream, which could contribute a noticeable value to the total operation cost.However, if TLS is enabled, then we could authenticate once (should be easy, but not trivial, think about replay attacks). Even disabling signature checking in trusted environment (localhost endpoint) can provide benefits, as the setup with s3/node being on one physical server is common.
Ideally, we could somehow extract FrostFS identity from the client certificate, but this is much more intrusive.
tree.GetOpLog
is another useful case, which can speed up tree synchronization.Another option is to make client responsible for disabling signatures
Though this won't work for client-side streams.
We could also significantly reduce CPU load by avoiding the need to sign each request-response in a series of HEAD requests during replication from the same node in the same TLS connection.