[SERVER-17251] with journal turned off journal compression option seems to affect throughput Created: 11/Feb/15 Updated: 08/Apr/15 Resolved: 22/Feb/15 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | WiredTiger |
| Affects Version/s: | 3.0.0-rc8 |
| Fix Version/s: | None |
| Type: | Bug | Priority: | Major - P3 |
| Reporter: | Asya Kamsky | Assignee: | Asya Kamsky |
| Resolution: | Cannot Reproduce | Votes: | 0 |
| Labels: | None | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Backwards Compatibility: | Fully Compatible |
| Operating System: | ALL |
| Participants: |
| Description |
|
Trying to test how much journal compression affects throughput with journal turned on, I inadvertently timed multiple runs with journal turned off both with and without journal compression type specified as none (vs snappy). Surprisingly I see a consistent difference, where none gives better performance. But journal compression setting should be irrelevant if journaling is disabled. |
| Comments |
| Comment by Asya Kamsky [ 22/Feb/15 ] |
|
seems to have been a statistical fluke, cannot reproduce. |
| Comment by Adam Comerford [ 11/Feb/15 ] |
|
Did some more testing, cut the threads in half on the client, added zlib for a couple of runs just to be sure. Pretty consistent, not seeing a differential now with my testing - entirely possible the variance I saw last night was CPU related as well (was running everything on the same host at that point): 4 threads, journal disabled, journal compressor = snappy run 1: 318.321 4 threads, journal disabled, journal compressor = none run 1: 321.594 4 threads, journal disabled, journal compressor = zlib run 1: 320.312 |
| Comment by Adam Comerford [ 11/Feb/15 ] |
|
EDIT: removing previous udpate because I suspect something was stealing CPU cycles randomly on my client machine and creating variance. I cleaned things up and got the following results: 8 threads, journal: disabled, journalCompressor: none run 1: 259.824s 8 threads, journal: disabled, journalCompressor: snappy run 1: 258.777s So, basically the same, at least for these tests where my client is the limiting factor (CPU bound). My server is not even breathing hard (<50% CPU, <50% IO) |