[SERVER-621] preallocation as an option Created: 10/Feb/10  Updated: 12/Jul/16  Resolved: 17/May/10

Status: Closed
Project: Core Server
Component/s: Performance
Affects Version/s: None
Fix Version/s: None

Type: Improvement Priority: Major - P3
Reporter: Tomasz Drobiszewski Assignee: Eliot Horowitz (Inactive)
Resolution: Done Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified
Environment:

Solaris ZFS others copy on write file systems


Participants:

 Description   

Preallocation kill performance on ZFS. It makes only bigger fragmentation, there is any performance benefits using pre allocation on ZFS, BTRFS ...
Before "fragmentation crash" MongoDB on compressed ZFS file system works 20%-30% faster than on UFS (reads).

I've tested it on Solaris, with 40G of RAM, 8CPU, 2.1T ZFS/UFS filesystem

1.8T of logs, 8 370 000 000 rows, seq writes and random reads



 Comments   
Comment by Eliot Horowitz (Inactive) [ 17/May/10 ]

we're now using posix_falocate, which should alleviate issues on zfs.
can you try a 1.5.x nightly?

Comment by Eliot Horowitz (Inactive) [ 10/Feb/10 ]

do you mean preallocation or filling with 0s?
you can turn off preallocation wtih --noprealloc but i suspect you mean filling with 0s
can you confirm?

Generated at Thu Feb 08 02:54:41 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.