HPE Storage Users Group
https://3parug.net/

P10400 array with vmfs datastores, should I still run defrag
https://3parug.net/viewtopic.php?f=18&t=529
Page 1 of 1

Author:  mujzeptu [ Mon Jan 06, 2014 8:34 pm ]
Post subject:  P10400 array with vmfs datastores, should I still run defrag

We have a V400 array running alll vmfs datastores with vcenter 5.1. We have hundreds of windows and linux servers, and I am wondering how helpful running local defrags within the OS can really be?

I have all vms running eager zeroed, with the 3par thin suite and all the VV's being thin. I am working on running vmfstools on the datastores to zero out unused space, I am making sure to compact CPGs after changes, tunevv, etc.

I am just curious how helpful OS defrags are on the individual vm OS's, and the reason/logic behind it.

Author:  mujzeptu [ Mon Jan 06, 2014 8:43 pm ]
Post subject:  Re: P10400 array with vmfs datastores, should I still run de

I have found:

http://blogs.vmware.com/vsphere/2011/09 ... st-os.html

http://kb.vmware.com/selfservice/micros ... Id=1006810

but I am just curious about 3par users real-world experiences...

Author:  Josh26 [ Mon Jan 06, 2014 9:49 pm ]
Post subject:  Re: P10400 array with vmfs datastores, should I still run de

3PAR aside, defrag on any form of RAID doesn't make sense. How do you align blocks continuously when that block is striped across several physical disks?

Honestly a lot of the desire to defragment is a hangover from "performance tools" a few years back that whose sales relied on this being viewed as a requirement. A bit like screensavers.

Author:  hdtvguy [ Tue Jan 07, 2014 7:28 am ]
Post subject:  Re: P10400 array with vmfs datastores, should I still run de

I am torn on this, as most arrays do some sort of prefetch and if the data is contiguous it should get a cache hit for the next block if the prefetch grabbed it, if it is truly the next block that is needed. If the data from the OS is highly fragmented then the "next block" surely will not have what the OS might be looking for next since in the OS the next piece of requested data is somewhere else. I also agree with some of the posts in the blog mentioned, the entire chain needs to be optimized otherwise the weak link becomes the bottle next.

Author:  Richard Siemers [ Tue Jan 07, 2014 2:05 pm ]
Post subject:  Re: P10400 array with vmfs datastores, should I still run de

Is there a problem that can be traced to a cause? Is it disk geometry based or are there issues with the file system allocation tables being larger due to excessive fragmentation? FAT32 vs NTFS, ETX2 vs EXT3, etc etc.

In modern datacenters, with modern file systems and modern SAN storage... I don't see or hear much, if any, concerns with file fragmentation.

Page 1 of 1 All times are UTC - 5 hours
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/