[SERVER-16602] Review WiredTiger default settings for engine and collections Created: 19/Dec/14  Updated: 15/Jan/15  Resolved: 13/Jan/15

Status: Closed
Project: Core Server
Component/s: Storage
Affects Version/s: 2.8.0-rc3
Fix Version/s: 2.8.0-rc5

Type: Improvement Priority: Major - P3
Reporter: Alexander Gorrod Assignee: Alexander Gorrod
Resolution: Done Votes: 0
Labels: wiredtiger
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Related
is related to SERVER-16269 WiredTiger blocks queries and updates... Closed
Backwards Compatibility: Fully Compatible
Participants:

 Description   

The current setting for memory_page_max in WiredTiger collections is 100MB. This helped with throughput on insert heavy workloads.

We have recently made enhancements to WiredTiger to lessen the benefit of a large memory_page_max setting. We should consider switching to the default setting (5MB).

There are cases where an application thread writes a page to disk after it reaches the memory_page_max setting - having this value large can cause operations to experience high latency.

I've run a few workloads locally and see equivalent performance with a smaller memory_page_max, and generally significantly smaller values in the profiling of slow operations.



 Comments   
Comment by Michael Cahill (Inactive) [ 13/Jan/15 ]

Check that documentation of WiredTiger defaults matches the new values, as described in the last update to this issue.

Comment by Githook User [ 13/Jan/15 ]

Author:

{u'username': u'michaelcahill', u'name': u'Michael Cahill', u'email': u'michael.cahill@wiredtiger.com'}

Message: SERVER-16602: Update WiredTiger default settings.

Comment by Eliot Horowitz (Inactive) [ 21/Dec/14 ]

Can we do a full mongo perf run?

Comment by Michael Cahill (Inactive) [ 19/Dec/14 ]

If the impact on multi-threaded insert workloads is minimal, I would support setting memory_page_max=10MB. I'm concerned about 5MB because inserting 5x 1MB documents will force eviction.

Can we gather the following data:

(1) impact on the throughput of a multi-threaded insert workload; and
(2) impact on latencies in a multi-threaded update workload.

Generated at Thu Feb 08 03:41:38 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.