[
https://issues.apache.org/jira/browse/CASSANDRA-1472?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Stu Hood updated CASSANDRA-1472:
--------------------------------
Attachment: 1472-v3.tgz
Attaching a third version of the patch. Since the last revision:
* 0001 through 0009 were updated to fix bugs, but didn't experience any
significant changes
* Integrated KEYS_BITMAP indexes into ColumnFamilyStore.scan (*0012*, *0013*)
** Added abstract db.secindex.SecondaryIndex, with KeysIndex and
KeysBitmapIndex subclasses
*** SecondaryIndex.certainty(expression) returns false if the index can return
false positives
*** SecondaryIndex.iterator(bounds, expression) iterates keys matching the index
** KeysBitmapIndex implements the probable key iteration by merging the results
from the sstable bitmap indexes with the results of a brute-force filtering of
all memtables (see TODOs)
* Indexes larger than memory are supported (*0010*)
** Bitmap indexes are recreated during usage of SSTableWriter (no change there)
** The 'io.sstable.bitidx.ScratchStack' object implements merging for bitidxs
which is very similar to SSTable compaction, but it occurs synchronously while
the SSTableWriter is in use:
### A segment of the index is built in memory, flushed to disk, and added to a
stack
### When enough segments of the same size exist, the stack merges them
** Although IO is still mostly sequential, SSTable write performance drops by
about 25% per index added. Haven't looked into bottlenecks here.
----
TODOs, in order of importance:
* Integrate rebuilding bitidxs into SSTable recovery
* Replace Memtable brute-force filtering with an in-memory index attached to
the Memtable
* Improve bin selection
----
I think this is probably ready for review if anyone has time: incorporating
KEYS_BITMAP indexes into SSTable recovery is probably the last blocker.
> Add bitmap secondary indexes
> ----------------------------
>
> Key: CASSANDRA-1472
> URL: https://issues.apache.org/jira/browse/CASSANDRA-1472
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Stu Hood
> Fix For: 0.7.0
>
> Attachments: 1472-v3.tgz
>
>
> Bitmap indexes are a very efficient structure for dealing with immutable
> data. We can take advantage of the fact that SSTables are immutable by
> attaching them directly to SSTables as a new component (supported by
> CASSANDRA-1471).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
