[
https://issues.apache.org/jira/browse/CASSANDRA-8630?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14711632#comment-14711632
]
Ariel Weisberg commented on CASSANDRA-8630:
-------------------------------------------
bq. We could consider changing this for compaction readers, or at least for
throttled readers (which amount to the same thing). There's no reason not to
read 64Kb at a time for compaction, since we know we'll want all of the data.
This would make me sleep a little better.
bq. Not sure if I found the comment you are referring to, you mean the
description of the class Input stream around a fixed ByteBuffer? I can change
it to Input stream around a single ByteBuffer, is this what you meant?
[I think it's already gone actually. Can't find it in the
diff.|https://github.com/apache/cassandra/compare/trunk...stef1927:8630-3.0#diff-0d7319a2a430b96865289eb87b136b32L25]
bq. The readers are potentially used in different threads...
Right so there needs to be a copy, but you don't need to copy the same state
every time you read. You can make an immutable copy once on write, and then
share that indefinitely. I think you are on the right track with the isCopy
flag, but maybe make it a field that is called immutableCopy or something, and
shared copy returns the same immutable view of the state every time. So if
immutableCopy is null then this State object is the immutable copy.
> Faster sequential IO (on compaction, streaming, etc)
> ----------------------------------------------------
>
> Key: CASSANDRA-8630
> URL: https://issues.apache.org/jira/browse/CASSANDRA-8630
> Project: Cassandra
> Issue Type: Improvement
> Components: Core, Tools
> Reporter: Oleg Anastasyev
> Assignee: Stefania
> Labels: compaction, performance
> Fix For: 3.x
>
> Attachments: 8630-FasterSequencialReadsAndWrites.txt, cpu_load.png,
> flight_recorder_001_files.tar.gz, flight_recorder_002_files.tar.gz,
> mmaped_uncomp_hotspot.png
>
>
> When node is doing a lot of sequencial IO (streaming, compacting, etc) a lot
> of CPU is lost in calls to RAF's int read() and DataOutputStream's write(int).
> This is because default implementations of readShort,readLong, etc as well as
> their matching write* are implemented with numerous calls of byte by byte
> read and write.
> This makes a lot of syscalls as well.
> A quick microbench shows than just reimplementation of these methods in
> either way gives 8x speed increase.
> A patch attached implements RandomAccessReader.read<Type> and
> SequencialWriter.write<Type> methods in more efficient way.
> I also eliminated some extra byte copies in CompositeType.split and
> ColumnNameHelper.maxComponents, which were on my profiler's hotspot method
> list during tests.
> A stress tests on my laptop show that this patch makes compaction 25-30%
> faster on uncompressed sstables and 15% faster for compressed ones.
> A deployment to production shows much less CPU load for compaction.
> (I attached a cpu load graph from one of our production, orange is niced CPU
> load - i.e. compaction; yellow is user - i.e. not compaction related tasks)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
