Details
-
Improvement
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
HA branch (HDFS-1623), 0.23.2
-
None
-
Reviewed
Description
One of the performance issues noticed in the HA branch is due to the much larger edit logs, now that we are writing OP_ADD transactions to the edit log on every block allocation. We can condense these calls down in two ways:
1) use variable-length integers for the block list length, size, and genstamp (most of these end up fitting in far less than 8 bytes)
2) use delta-coding for the genstamp and block size for any blocks after the first block (most blocks will be the same size and only slightly higher genstamps)
3) introduce a new OP_UPDATE_BLOCKS transaction that doesn't re-serialize metadata information like lease owner, permissions, etc
4) allow OP_UPDATE_BLOCKS to only re-serialize the blocks that have changed for a given transaction
Attachments
Attachments
Issue Links
- is related to
-
HDFS-3010 Get performance on HA branch to match trunk
- Resolved
Apparently I am bad at counting. The above should read "in four ways"
I ran a teragen with 580 mappers and a 4MB block size on the HA branch with optimizations 1 and 2 in place and it reduced the size of the edit log from 1.2GB to ~600MB (factor of two savings). This benchmark is designed to log a lot of persistBlocks() calls, so the savings in less strenuous use cases won't be as large, but still worth doing.