Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[TRACKING ISSUE] TPC default to 1 connection #1211

Open
hz-devops-test opened this issue Nov 7, 2023 · 0 comments
Open

[TRACKING ISSUE] TPC default to 1 connection #1211

hz-devops-test opened this issue Nov 7, 2023 · 0 comments

Comments

@hz-devops-test
Copy link

The tracking issue for the Java side PR.

See hazelcast/hazelcast#25871 for details.


Before this PR, when a TPC aware client connects to a TPC enabled server, the server sends the list of TPC ports it can connect to. So if the server has 64 TPC threads, it will open 64 server sockets and the client will make 64 connections.

The problem is that having many connections, leads to smaller and more packets which causes additional processing overhead and can run into packets per seconds limits (e.g. on AWS). And the performance penalty of the PPS limit is a lot higher than the penalty of context switching.

This PR changes that. By default, a client will now connect to 1 random port from the TPC port list. The Hazelcast member will then route the request to the right core (the same as with classic networking). The number of ports can be configured through the ClientTpcConfig object and one can restore the behavior of connecting to all TPC ports or a subset.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant