[SERVER-17538] Mongodb 3.0 with wiredtiger has poor read/aggregate performance Created: 11/Mar/15  Updated: 28/Apr/15  Resolved: 28/Apr/15

Status: Closed
Project: Core Server
Component/s: Aggregation Framework, Performance, WiredTiger
Affects Version/s: 3.0.0
Fix Version/s: None

Type: Question Priority: Minor - P4
Reporter: Sundarajan Srinivasan Assignee: Ramon Fernandez Marina
Resolution: Done Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Related
Participants:

 Description   

Performance comparison for MMAPv1 and WiredTiger

1. Write performance is good when using wiredtiger as a storage engine.
2. But read performance is very poor compared to MMAPv1 engine.

Hardware Configuration: Ubuntu 12.04
CPU Cores: 2
RAM: 8GB

We ran scenarios with both MMAP and wiredtiger using the same above hardware configurations. Now we finally are in a position to conclude that MMAP gives us a better performance compared to WiredTiger engine.

Help would be appreciated.

Is there any specific configuration params available to increase the read performance.



 Comments   
Comment by Ramon Fernandez Marina [ 28/Apr/15 ]

sundarajan.s@gmail.com, from your description I don't see evidence of a bug in the server. Different storage engines perform better in different areas, and MMAPv1 may give you better read performance than WiredTiger in some cases. Since the SERVER project is for reporting bugs or feature suggestions for the MongoDB server I'm going to resolve this ticket. We run performance tests internally, and if we see any results that match this behavior and point to a bug we can re-open this ticket.

For further MongoDB-related support discussion I'd recommend you post on the mongodb-user group or Stack Overflow with the mongodb tag, where your question will reach a larger audience. A question like this involving more discussion would be best posted on the mongodb-user group.

Regards,
Ramón.

Comment by Sundarajan Srinivasan [ 12/Mar/15 ]

Please find the below information for your perusal.

1. MongoDB is present in clustered environment (2 mongod + configserver + mongos) with wiredtiger enabled.
2. Concurrently 150 users were writing 100000 documents and reading 80000 documents in total.
3. I can able to say that all the writes were really good from the inference made from the system.profile collection.
4. But we are using aggregation queries that really takes more time. Say like 4 to 5s for a request which is really costly.
5. CPU utilization was maximum for 2 core used.
6. We are using a AWS Ubuntu server.
7. One last thing is, I saw a huge amoun of receive queue formation were the mongodb is mounted.

Ubuntu Version - 12.04
RAM: 8 GB

I am not sure on the storage type as we are using AWS version and dont think i can share the software details we are using.

Any help to optimize would be really appreciated.

Thanks
Sundar

Comment by Ramon Fernandez Marina [ 11/Mar/15 ]

I'm afraid this is not enough information to determine whether there's a performance-related bug in WiredTiger read-heavy loads. Can you please elaborate on what kind of tests did you run? Did you use any common tools like YCSB or iibench? If you used your own software, can you please share it with us so we can track down any potential performance bottlenecks?

It would be useful to know what kind of storage you used (spinning disks, SSDs, etc.) and what kind of setup did you do your testing on (stand-alone vs. replica sets, sharded vs. non-sharded).

Thanks,
Ramón.

Generated at Thu Feb 08 03:44:49 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.