[SERVER-2097] partitioning Created: 15/Nov/10  Updated: 09/Feb/23

Status: Backlog
Project: Core Server
Component/s: Performance, Storage
Affects Version/s: None
Fix Version/s: None

Type: New Feature Priority: Major - P3
Reporter: Markus Gattol Assignee: Michael Gargiulo
Resolution: Unresolved Votes: 24
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Related
Participants:
Case:

 Description   

Like currently in existence RDBMSs provide so-called
partitioning, it would also be nice to have this with MongoDB.

If not partitioning then at least something akin like clustering
similar data on disk in order to reduce seeks (might become less
of an issue the more SSD penetrates the storage market).



 Comments   
Comment by Wernfried Domscheit [ 16/Apr/21 ]

I think this can be done with Zone Sharding

Comment by Dmitry Matz [ 11/Jun/19 ]

Hello Team!

We're also very interesting in this feature as we're going to keep historical data in MongoDB and want to collect them on a daily basis.

Could you please share your vision and plans regarding this feature?

Comment by Tatiana [ 05/Jul/18 ]

Hi guys, we also would like mongo to be able to partition collections. Any idea about the eta ? Is it on the roadmap at all? 

 

Thanks!

Comment by Giri Saranu [ 11/Aug/17 ]

We are looking for partitioning at collection due to the size of the our collection data size. If the partitioning feature doesn't support in Mongo, we might have to look into some other databases like Postgres.

Comment by andy [ 28/Oct/15 ]

This feature is a must-have for us. We are collecting time-based data (lots of it) in single node installations and we want user to have control over data life span (how old data is retained). Ordinary remove() is too slow for us.

Two currently available solutions in MongoDB are TTL and capped collections, neither of which are appropriate in our case:

  • TTL still removes record-by-record. While app doesn't have to do that, MongoDB server still suffers performance problems. Also, we want the user to be able to change this time whenever he wishes.
  • capped collections don't give any guarantee as to the time frame that will be retained.

This is a good overview of options available: https://dzone.com/articles/storing-time-series-data

Compression (which was not supported back then by MongoDB) and partitioned collections were the reasons we migrated to TokuMX: https://www.percona.com/blog/2014/05/29/introducing-partitioned-collections-for-mongodb-applications/
However since being acquired by Percona there was little activity on their front so we will probably have to migrate back to MongoDB (if we can) or elsewhere.

When making such system, it would be nice if one could convert existing tables to partitioned ones without migration. TokuMX doesn't allow that (you need to create new collection).

Do you have any plans on supporting partitioned collections?

Comment by Eliot Horowitz (Inactive) [ 15/Nov/10 ]

partitioning is a very specific thing

Generated at Thu Feb 08 02:58:58 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.