Thick volume: no negative performance impact other than temporally snapshot use space right after running defrag -> OK to run if fragmentation is high on Windows
Thin volume: may reduce free space on Equallogic -> Don't run until after running test even fragmentation is high (May need to contact Dell and VMware for official answer).
As Don mentions, this is a work/gain consideration.
Defraging systems (and guests) on the SAN does ensure system optimizations like read ahead for sequential (backups) and total I/Os done to complete work are optimized.
This comes at the price of you do more work while defragmenting. It also has temporary side effects that near term snapshots, and replication consume more disk and network bandwidth.
So it is a tradeoff of work vs. gain. Never doing it says you minimized work, but things get less efficient more and more over time.
Doing it too frequently says you made defragmenting a workload unto itself that has a permanent ongoing cost. I agree with Don's caution to too frequently is a bad thing.
I suggest you monitor fragementation, and once every 6 months or so (or after you know you have made a set of changes and will be done for a quarter or more), check the fragmentation and defragment the ones that show a lot of fragmentation.
This also applies to shrinking of virtual machines - it is the same work/gain tradeoff.
Dyna3
2 Posts
0
September 13th, 2012 08:00
Thank you for your answer.
So to summarize this:
Thick volume: no negative performance impact other than temporally snapshot use space right after running defrag -> OK to run if fragmentation is high on Windows
Thin volume: may reduce free space on Equallogic -> Don't run until after running test even fragmentation is high (May need to contact Dell and VMware for official answer).
eschott
1 Message
0
September 13th, 2012 11:00
As Don mentions, this is a work/gain consideration.
Defraging systems (and guests) on the SAN does ensure system optimizations like read ahead for sequential (backups) and total I/Os done to complete work are optimized.
This comes at the price of you do more work while defragmenting. It also has temporary side effects that near term snapshots, and replication consume more disk and network bandwidth.
So it is a tradeoff of work vs. gain. Never doing it says you minimized work, but things get less efficient more and more over time.
Doing it too frequently says you made defragmenting a workload unto itself that has a permanent ongoing cost. I agree with Don's caution to too frequently is a bad thing.
I suggest you monitor fragementation, and once every 6 months or so (or after you know you have made a set of changes and will be done for a quarter or more), check the fragmentation and defragment the ones that show a lot of fragmentation.
This also applies to shrinking of virtual machines - it is the same work/gain tradeoff.