[SERVER-20596] Performance regression in new mongos query path (about 17% worse than 3.0) Created: 24/Sep/15 Updated: 06/Dec/22 Resolved: 24/Oct/19 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | Performance, Querying, Sharding |
| Affects Version/s: | 3.1.9 |
| Fix Version/s: | None |
| Type: | Bug | Priority: | Major - P3 |
| Reporter: | Rui Zhang (Inactive) | Assignee: | [DO NOT USE] Backlog - Sharding Team |
| Resolution: | Incomplete | Votes: | 2 |
| Labels: | sys-perf-reg | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Attachments: |
|
||||||||||||||||||||||||||||||||
| Issue Links: |
|
||||||||||||||||||||||||||||||||
| Assigned Teams: |
Sharding
|
||||||||||||||||||||||||||||||||
| Operating System: | ALL | ||||||||||||||||||||||||||||||||
| Sprint: | QuInt B (11/02/15), QuInt C (11/23/15) | ||||||||||||||||||||||||||||||||
| Participants: | |||||||||||||||||||||||||||||||||
| Case: | (copied to CRM) | ||||||||||||||||||||||||||||||||
| Linked BF Score: | 0 | ||||||||||||||||||||||||||||||||
| Description |
|
During the 3.1 development cycle, the implementation of find and getMore on mongos was rewritten under SERVER-15176. This new mongos query path plugs into the TaskExecutor event loop framework. It issues find, getMore, and killCursors commands to the shards using the new asynchronous networking layer. Our automated industry benchmarks performance loop shows that the new query path is about 17% slower than its 3.0 predecessor. This ticket tracks the work to achieve parity with 3.0 on the industry benchmark workload, which could involve optimizations to the asynchronous networking code, the task executor layer, or the new query path itself. Original DescriptionRead performance regression found in system-perf EVG test. This is due to following change: this enable new mongos query path for legacy read, therefore, here we are showing the performance issue with new path. A single findOne test with a single sharded cluster with benchRun results for b1982b:
32e55, the parent SHA
|
| Comments |
| Comment by Sheeri Cabral (Inactive) [ 24/Oct/19 ] |
|
Closing, tests showed it wasn't query path related, and the purported regression is from a years-old version. |
| Comment by David Storch [ 23/Nov/15 ] |
|
We've made substantial performance improvements under the various linked tickets. However, we have not yet achieved parity with version 3.0 on the industry benchmarks workload. I have updated the title and description of this ticket to reflect our progress. Bumping this ticket into 3.3 Required, since the current perf is sufficient for the 3.2.0 release. |
| Comment by Githook User [ 09/Nov/15 ] |
|
Author: {u'username': u'andy10gen', u'name': u'Andy Schwerin', u'email': u'schwerin@mongodb.com'}Message: |
| Comment by Githook User [ 05/Nov/15 ] |
|
Author: {u'username': u'dstorch', u'name': u'David Storch', u'email': u'david.storch@10gen.com'}Message: |
| Comment by Githook User [ 22/Oct/15 ] |
|
Author: {u'username': u'dstorch', u'name': u'David Storch', u'email': u'david.storch@10gen.com'}Message: |