hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Nick Dimiduk (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HBASE-11331) [blockcache] lazy block decompression
Date Wed, 06 Aug 2014 19:36:12 GMT

     [ https://issues.apache.org/jira/browse/HBASE-11331?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Nick Dimiduk updated HBASE-11331:

    Attachment: lazy-decompress.02.1.pdf

re: lazy-decompress.02.1.pdf

Attaching some metrics after running v02 with less os page cache and feature enabled/disabled.

This time the total RS heap is increased to 20g, so the effective blockcache size is ~12g.
The machine has 24g of RAM, so this squeezes the OS-available memory down to ~4g -- when the
blockcache is full. I've added blockcache size and OS page cache to the charts  Same table,
same compression ratio, same PE randomRead test, this time run with --size=45, so I expect
~11g of compressed blocks.

I expected this to be an IO-bound workload with flag=false but to be cache-at-capacity for
flag=true. The massive evictions look to me like the blockcache heapsize calculation is based
on the decompressed size, not compressed size. Updating patch.

> [blockcache] lazy block decompression
> -------------------------------------
>                 Key: HBASE-11331
>                 URL: https://issues.apache.org/jira/browse/HBASE-11331
>             Project: HBase
>          Issue Type: Improvement
>          Components: regionserver
>            Reporter: Nick Dimiduk
>            Assignee: Nick Dimiduk
>         Attachments: HBASE-11331.00.patch, HBASE-11331.01.patch, HBASE-11331.02.patch,
HBASE-11331LazyBlockDecompressperfcompare.pdf, lazy-decompress.02.0.pdf, lazy-decompress.02.1.pdf
> Maintaining data in its compressed form in the block cache will greatly increase our
effective blockcache size and should show a meaning improvement in cache hit rates in well
designed applications. The idea here is to lazily decompress/decrypt blocks when they're consumed,
rather than as soon as they're pulled off of disk.
> This is related to but less invasive than HBASE-8894.

This message was sent by Atlassian JIRA

View raw message