cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Joshua McKenzie (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-7282) Faster Memtable map
Date Wed, 20 Aug 2014 13:32:27 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-7282?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14103866#comment-14103866
] 

Joshua McKenzie commented on CASSANDRA-7282:
--------------------------------------------

Some interesting results on the [99.9th|http://riptano.github.io/cassandra_performance/graph_v3/graph.html?stats=stats.7282.defaults.json&metric=99.9th_latency&operation=1_write&smoothing=1&show_aggregates=true&xmin=0&xmax=688.16&ymin=0&ymax=1600]
and [max|http://riptano.github.io/cassandra_performance/graph_v3/graph.html?stats=stats.7282.defaults.json&metric=max_latency&operation=1_write&smoothing=1&show_aggregates=true&xmin=0&xmax=688.16&ymin=0&ymax=3000]
on write latency.  While their aggregated values look better on the new non-blocking hashordered
map there look to be more outliers.  Any ideas on why that might be [~benedict]?  95th and
99th shows pretty [marked improvement|http://riptano.github.io/cassandra_performance/graph_v3/graph.html?stats=stats.7282.defaults.json&metric=99th_latency&operation=1_write&smoothing=1&show_aggregates=true&xmin=0&xmax=688.16&ymin=0&ymax=140]
so the differential looks to be on the extremes only.

Either way - I'll be very interested to see the results of the test w/memtable and row size
changes.

> Faster Memtable map
> -------------------
>
>                 Key: CASSANDRA-7282
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-7282
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: Benedict
>            Assignee: Benedict
>              Labels: performance
>             Fix For: 3.0
>
>
> Currently we maintain a ConcurrentSkipLastMap of DecoratedKey -> Partition in our
memtables. Maintaining this is an O(lg(n)) operation; since the vast majority of users use
a hash partitioner, it occurs to me we could maintain a hybrid ordered list / hash map. The
list would impose the normal order on the collection, but a hash index would live alongside
as part of the same data structure, simply mapping into the list and permitting O(1) lookups
and inserts.
> I've chosen to implement this initial version as a linked-list node per item, but we
can optimise this in future by storing fatter nodes that permit a cache-line's worth of hashes
to be checked at once,  further reducing the constant factor costs for lookups.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message