[
https://issues.apache.org/jira/browse/CASSANDRA-6127?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13820499#comment-13820499
]
Quentin Conner edited comment on CASSANDRA-6127 at 11/12/13 9:34 PM:
---------------------------------------------------------------------
The num_tokens setting has a mild impact on average cpu utilization. Please
see the graph below for the trend with 256 nodes.
!cpu-vs-token-graph.png!
This graph does not characterize the bursty nature of any given node's CPU
utilization. It does average the utilization over a 200 second period, taken
at 10 second intervals using "sar -u".
Since gossip heartbeat destinations are random, "unlucky" nodes will sometimes
receive twice the gossip traffic and (10 second basis) CPU utilization has been
casually observed @25% for N=256 and 60% for N=512.
was (Author: qconner):
The num_tokens setting has a mild impact on average cpu utilization. Please
see the graph for the trend with N=256, 256 nodes.
This graph does not characterize the bursty nature of any given node's CPU
utilization. It does average the utilization over a 200 second period, taken
at 10 second intervals using "sar -u".
Since gossip heartbeat destinations are random, "unlucky" nodes will sometimes
receive twice the gossip traffic and (10 second basis) CPU utilization has been
casually observed @25% for N=256 and 60% for N=512.
> vnodes don't scale to hundreds of nodes
> ---------------------------------------
>
> Key: CASSANDRA-6127
> URL: https://issues.apache.org/jira/browse/CASSANDRA-6127
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Environment: Any cluster that has vnodes and consists of hundreds of
> physical nodes.
> Reporter: Tupshin Harper
> Assignee: Jonathan Ellis
> Attachments: 2013-11-05_18-04-03_no_compression_cpu_time.png,
> 2013-11-05_18-09-38_compression_on_cpu_time.png, 6000vnodes.patch,
> AdjustableGossipPeriod.patch, cpu-vs-token-graph.png,
> delayEstimatorUntilStatisticallyValid.patch, flaps-vs-tokens.png
>
>
> There are a lot of gossip-related issues related to very wide clusters that
> also have vnodes enabled. Let's use this ticket as a master in case there are
> sub-tickets.
> The most obvious symptom I've seen is with 1000 nodes in EC2 with m1.xlarge
> instances. Each node configured with 32 vnodes.
> Without vnodes, cluster spins up fine and is ready to handle requests within
> 30 minutes or less.
> With vnodes, nodes are reporting constant up/down flapping messages with no
> external load on the cluster. After a couple of hours, they were still
> flapping, had very high cpu load, and the cluster never looked like it was
> going to stabilize or be useful for traffic.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
