[
https://issues.apache.org/jira/browse/CASSANDRA-1608?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13040541#comment-13040541
]
Benjamin Coverston edited comment on CASSANDRA-1608 at 5/28/11 5:11 PM:
------------------------------------------------------------------------
There's probably nothing that prevents us from doing that. Is our goal here to
replace compaction entirely?
The manifest information consists, minimally of the level information and
ranges. For us ranges are easy as they are readily available when the SSTables
are read in at restart, flushing, or compaction.
Taking a stab at this I made the compaction manager abstract, then created a
concrete implementation for the current compaction implementation. Happily
hacking on a level based compaction manager I kept running into a delemma:
Where do I store the level information. There are a few options:
1. The descriptor A hack, simple, but also adds information that probably
wouldn't be used by any other compaction manager, yet it would be there. Unless
we're moving head-long into a level-db approach I'm not super excited about
this.
2. Store it on a per-sstable basis -in- the sstable: To continue along this
path I would like to have a standard place to put "extra" metadata in the
sstables. A header of some sort. I like the idea of using a metadata block in
the SSTables to store this type of information.
3. Use an on-disk manifest. -- Pro: only my compaction manager needs to deal
with this information, but there is a non-trivial amount of bookeeping that
would need to be done to ensure this is kept up to day and valid.
EDIT:
4. This is probably the best option, create a new component type:
METADATA_STORE which will hold namespaced key/value pairs on a per-sstable
basis.
was (Author: bcoverston):
There's probably nothing that prevents us from doing that. Is our goal here
to replace compaction entirely?
The manifest information consists, minimally of the level information and
ranges. For us ranges are easy as they are readily available when the SSTables
are read in at restart, flushing, or compaction.
Taking a stab at this I made the compaction manager abstract, then created a
concrete implementation for the current compaction implementation. Happily
hacking on a level based compaction manager I kept running into a delemma:
Where do I store the level information. There are a few options:
1. The descriptor A hack, simple, but also adds information that probably
wouldn't be used by any other compaction manager, yet it would be there. Unless
we're moving head-long into a level-db approach I'm not super excited about
this.
2. Store it on a per-sstable basis -in- the sstable: To continue along this
path I would like to have a standard place to put "extra" metadata in the
sstables. A header of some sort. I like the idea of using a metadata block in
the SSTables to store this type of information.
3. Use an on-disk manifest. -- Pro: only my compaction manager needs to deal
with this information, but there is a non-trivial amount of bookeeping that
would need to be done to ensure this is kept up to day and valid.
> Redesigned Compaction
> ---------------------
>
> Key: CASSANDRA-1608
> URL: https://issues.apache.org/jira/browse/CASSANDRA-1608
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Chris Goffinet
>
> After seeing the I/O issues in CASSANDRA-1470, I've been doing some more
> thinking on this subject that I wanted to lay out.
> I propose we redo the concept of how compaction works in Cassandra. At the
> moment, compaction is kicked off based on a write access pattern, not read
> access pattern. In most cases, you want the opposite. You want to be able to
> track how well each SSTable is performing in the system. If we were to keep
> statistics in-memory of each SSTable, prioritize them based on most accessed,
> and bloom filter hit/miss ratios, we could intelligently group sstables that
> are being read most often and schedule them for compaction. We could also
> schedule lower priority maintenance on SSTable's not often accessed.
> I also propose we limit the size of each SSTable to a fix sized, that gives
> us the ability to better utilize our bloom filters in a predictable manner.
> At the moment after a certain size, the bloom filters become less reliable.
> This would also allow us to group data most accessed. Currently the size of
> an SSTable can grow to a point where large portions of the data might not
> actually be accessed as often.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
