[
https://jira.jboss.org/jira/browse/JGRP-685?page=com.atlassian.jira.plugi...
]
Bela Ban commented on JGRP-685:
-------------------------------
For a cluster of N nodes, we should get Bandwidth / (N -1) data per second, e.g. with a
GBit switch and 4 nodes, we should get 1 GBit / 3 == 333 MBit/sec == 41MB/sec.
Verify this with perf.Test and perf.IPerf
Perf test: mimick buddy replication
-----------------------------------
Key: JGRP-685
URL:
https://jira.jboss.org/jira/browse/JGRP-685
Project: JGroups
Issue Type: Task
Reporter: Bela Ban
Assignee: Bela Ban
Fix For: 2.7
Things to do with the perf test:
- Mimick buddy replication: send messages not to self, but only to a selected buddy
- Measure time to send 100MB to A, B, C, D
- New look at HTOTAL, might help with perf in this scenario
- TcpTransport (95MB/sec) versus JGroupsTransport/tcp.xml (54MB/sec): why the diff ?
- Members: A (send), B (receive)
- Might get better perf with impl of first bullet item (above)
- JPerf (between A and B) gets 113MB/sec !
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
https://jira.jboss.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see:
http://www.atlassian.com/software/jira