[CASSANDRA-6386] FD mean calculation performance improvement

• Type: Improvement
• Status: Resolved
• Priority: Low
• Resolution: Fixed
• Fix Version/s: 1.2.13, 2.0.3
• Assignee: Jonathan Ellis
• Reporter: Quentin Conner

Authors

Jonathan Ellis

Reviewers

Brandon Williams

• Created: 20/Nov/13 16:57
• Updated: 16/Apr/19 09:31
• Resolved: 20/Nov/13 18:47

Description

CPU profiling in #6127 found the o.a.c.gms.arrrivalWindow.mean() calculation taking the most cpu time during steady state gossip.

cpu usage can be a limiting factor when starting > 500 nodes concurrently, which customers sometimes do (though we suggest starting nodes 2 or 3 minutes apart).

Issue Links

• blocks: CASSANDRA-6127 vnodes don't scale to hundreds of nodes