[
https://issues.apache.org/jira/browse/CASSANDRA-5455?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13666888#comment-13666888
]
Peter Bailis commented on CASSANDRA-5455:
-----------------------------------------
Okay. #1 will likely require more extensive code changes: basically, it'll
require EstimatedHistograms for each of the servers acting as replicas for a
given ColumnFamily and will require EstimatedHistogram tracing in the
StorageProxy (to separate network-based latency from disk-based latency). Are
these changes feasible?
re: "a window of individual latency times," looking at the Metrics
implementation of EstimatedHistogram, EstimatedHistogram.values() should
provide a reasonable enough sample (especially since, as you mention, since it
has other uses as well).
Perhaps the simplest strategy is to go with #3 for now but implement #1 in the
future if there's interest. #3 is easy; I've already written an example
external module to do RTT/2 predictions:
https://github.com/pbailis/pbs-predictor/blob/9d31acd1667b08affa609278689b540d8e0380f5/pbspredictor/src/main/java/edu/berkeley/pbs/cassandra/CassandraLatencyTrace.java
> Remove PBSPredictor
> -------------------
>
> Key: CASSANDRA-5455
> URL: https://issues.apache.org/jira/browse/CASSANDRA-5455
> Project: Cassandra
> Issue Type: Bug
> Reporter: Jonathan Ellis
> Assignee: Jonathan Ellis
> Fix For: 2.0
>
> Attachments: 5455.txt
>
>
> It was a fun experiment, but it's unmaintained and the bar to understanding
> what is going on is high. Case in point: PBSTest has been failing
> intermittently for some time now, possibly even since it was created. Or
> possibly not and it was a regression from a refactoring we did. Who knows?
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
