| Bug #54563 | InnoDB: Performance drop in varios sysbench tests | ||
|---|---|---|---|
| Submitted: | 16 Jun 2010 19:19 | Modified: | 17 Jun 2010 8:06 |
| Reporter: | Alexey Stroganov | Email Updates: | |
| Status: | Closed | Impact on me: | |
| Category: | MySQL Server: InnoDB storage engine | Severity: | S3 (Non-critical) |
| Version: | 5.5.5pre | OS: | Any |
| Assigned to: | Sunny Bains | CPU Architecture: | Any |
[17 Jun 2010 1:11]
Sunny Bains
This was an incomplete revert of 3008.2.76..3008.2.78. I've reverted one other change that was not reverted earlier.
[17 Jun 2010 8:06]
Alexey Stroganov
Exact changes that caused performance regression were identified and reverted. Results are now on par with 5.5.4:
#
# Test: OLTP_RO:throughput, TPS
#
# Server 1 - 5.5.4-m3
# Server 2 - 5.5.5-m3-pre-fixed
#
# Server 1 Server 2
# Thread INNODB INNODB
4 1738.73 1745.53
16 5309.76 5162.01
64 7280.48 7308.89
128 7127.66 7170.85
256 6873.85 6972.87
Description: While running sysbench tests for the current mysql-trunk branch we have observing performance drops in various sysbench tests for InnoDB. Most notable drop was observed in OLTP_RO: # Test: OLTP_RO:throughput, TPS # # Server 1 - 5.5.4-m3 # Server 2 - 5.5.5-m3-pre-trunk.runtime.110610 # # Server 1 Server 2 # Thread INNODB INNODB 4 1738.73 1581.36 16 5309.76 5097.22 64 7280.48 6270.42 128 7127.66 6045.14 256 6873.85 5945.02 After retesting of several binaries we've managed to isolate issue to following revisions: 3008.2.78 - performance drop 3008.2.77 - crashes 3008.2.76 - does not compile 3008.2.75 - results are ok Below results for binary where changesets: 3008.2.76,3008.2.77,3008.2.78 were reverted: # Test: OLTP_RO:throughput, TPS # # Server 1 - 5.5.4-m3 # Server 2 - 5.5.5-m3-pre-runtime.without_rev.3008.2.76-78 # # Server 1 Server 2 # Thread INNODB INNODB 4 1738.73 1671.85 16 5309.76 5147.09 64 7280.48 7217.07 128 7127.66 6967.69 256 6873.85 6898.62 How to repeat: Run sysbench test for 5.5.5pre