[
https://issues.apache.org/jira/browse/HBASE-4218?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13188294#comment-13188294
]
Matt Corgan commented on HBASE-4218:
------------------------------------
I used plenty of memory and a warmup run so that for the measured results all
reads were served out of the OS page cache and HBase block cache. I was trying
to measure compression ratio and cpu performance assuming that the data set is
very hot and cached nearly 100%. If you're IO bound, then that 50% cpu
difference shouldn't matter much, like you said. It strikes me that bringing
IO into the test is really just testing the effective size of the block cache
which you can already do by adjusting the block cache size in hbase-site. CPU
efficiency difference would get drowned out.
A scenario i have where data is 100% cached is chronological log ("event") data
(sharded 16 ways) where the last ~2 days fit in memory. We add different
secondary index tables to the primary table depending on different reports we
want to generate. When scanning those secondary indexes we pull millions of
rows from the primary table in random order. The better the compression, the
more days of log events we can hold in memory, and the better the cpu
efficiency, the faster we can do the random reads.
> Data Block Encoding of KeyValues (aka delta encoding / prefix compression)
> ---------------------------------------------------------------------------
>
> Key: HBASE-4218
> URL: https://issues.apache.org/jira/browse/HBASE-4218
> Project: HBase
> Issue Type: Improvement
> Components: io
> Affects Versions: 0.94.0
> Reporter: Jacek Migdal
> Assignee: Mikhail Bautin
> Labels: compression
> Fix For: 0.94.0
>
> Attachments: 0001-Delta-encoding-fixed-encoded-scanners.patch,
> 0001-Delta-encoding.patch, 4218-2012-01-14.txt, 4218-v16.txt, 4218.txt,
> D447.1.patch, D447.10.patch, D447.11.patch, D447.12.patch, D447.13.patch,
> D447.14.patch, D447.15.patch, D447.16.patch, D447.17.patch, D447.18.patch,
> D447.19.patch, D447.2.patch, D447.20.patch, D447.21.patch, D447.22.patch,
> D447.23.patch, D447.24.patch, D447.3.patch, D447.4.patch, D447.5.patch,
> D447.6.patch, D447.7.patch, D447.8.patch, D447.9.patch,
> Data-block-encoding-2011-12-23.patch,
> Delta-encoding-2012-01-17_11_09_09.patch,
> Delta-encoding.patch-2011-12-22_11_52_07.patch,
> Delta-encoding.patch-2012-01-05_15_16_43.patch,
> Delta-encoding.patch-2012-01-05_16_31_44.patch,
> Delta-encoding.patch-2012-01-05_16_31_44_copy.patch,
> Delta-encoding.patch-2012-01-05_18_50_47.patch,
> Delta-encoding.patch-2012-01-07_14_12_48.patch,
> Delta-encoding.patch-2012-01-13_12_20_07.patch,
> Delta_encoding_with_memstore_TS.patch, open-source.diff
>
>
> A compression for keys. Keys are sorted in HFile and they are usually very
> similar. Because of that, it is possible to design better compression than
> general purpose algorithms,
> It is an additional step designed to be used in memory. It aims to save
> memory in cache as well as speeding seeks within HFileBlocks. It should
> improve performance a lot, if key lengths are larger than value lengths. For
> example, it makes a lot of sense to use it when value is a counter.
> Initial tests on real data (key length = ~ 90 bytes , value length = 8 bytes)
> shows that I could achieve decent level of compression:
> key compression ratio: 92%
> total compression ratio: 85%
> LZO on the same data: 85%
> LZO after delta encoding: 91%
> While having much better performance (20-80% faster decompression ratio than
> LZO). Moreover, it should allow far more efficient seeking which should
> improve performance a bit.
> It seems that a simple compression algorithms are good enough. Most of the
> savings are due to prefix compression, int128 encoding, timestamp diffs and
> bitfields to avoid duplication. That way, comparisons of compressed data can
> be much faster than a byte comparator (thanks to prefix compression and
> bitfields).
> In order to implement it in HBase two important changes in design will be
> needed:
> -solidify interface to HFileBlock / HFileReader Scanner to provide seeking
> and iterating; access to uncompressed buffer in HFileBlock will have bad
> performance
> -extend comparators to support comparison assuming that N first bytes are
> equal (or some fields are equal)
> Link to a discussion about something similar:
> http://search-hadoop.com/m/5aqGXJEnaD1/hbase+windows&subj=Re+prefix+compression
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
