Unsolved
This post is more than 5 years old
9 Posts
0
1350
How to balance backend disks
I have a VMAX which has three disk groups (FC, SATA and EFD). I have three storage tiers defined/created for fast vp on the box; one for each type of disk. I have three fast policies defined; "performance" (which is 10% EFD, 40% FC and 100% SATA), "standard" (which is 4% EFD, 36% FC and 70% SATA) and "value (which is 10% FC and 100% SATA). All new writes goes to FC and fast make them move up or down based on policy.
During a code upgrade, EMC by accident, disabled FAST all together and we didn't notice that till few months.
FAST is enabled now and supposedly has caught up. But I am still seeing ~80% of my backend load on SATA. In other words, my backend disks don't look balanced at all.
What and how I can balance my backend disks?
sauravrohilla1
3 Posts
0
March 19th, 2014 21:00
The performance & value policies states that your data can remain 100% in SATA. Perhaps, data wants to remain in SATA only and it sees no benefit in promoting itself?
Another reason could be the higher Fast relocation rate value, if its set towards 10 then the rate at which data moves will be slower and you have also said that fast controller was disabled for a long time. May be there is loads of data to be moved around and it will slowly become balanced?
regards,
Saurabh Rohilla
rawstorage
419 Posts
0
March 23rd, 2014 05:00
Xubair, if you still have 80% of your load being directed at your sata tier, then your FAST VP policies may need to be examined.
If you are sharing the same underlying diskgroups/pools with all of your FAST VP tiers then you may be just redirecting the workload to SATA with your value and standard policies. If your storage groups that are associated with these policies are driving the majoirity of your workload, you may be trapping these loads on the SATA tier by restricting the FC/EFD usage.
Have a look at the storage groups driving your load and determine which is generating the majority of the workload, you may have some groups which aren't suited to the restrictive policies you have set here. especially the 10% FC 100% SATA. Host IO controls may be something to look at if if this is the case. Or perhaps a simpler approach ...
I suggest reviewing the guidelines set out in the TechNote
FAST VP for Symmetrix VMAX Theory and Best Practices for Planning and Performance
EMC's own Sean Cummins has an excellent blog that might explain a few things too..
VMAX FASTVP Best Practice Essentials | Sean Cummins
Quincy561
1.3K Posts
0
March 24th, 2014 10:00
If you have space in the FC tier, you could find some of the busiest devices/SGs and do a VLUN migration of them to the FC pool. Or VLUN those tracks in the SATA pool to the FC pool if your ucode is high enough to support this option.
donlab1
6 Posts
0
March 24th, 2014 10:00
what is output of these commands?
symfast list -control_parms -sid xxx
symfast -sid xxxx list -fp
symfast -sid xxxx list -state
donlab1
6 Posts
0
March 24th, 2014 10:00
what are your fast policies set to?
we use 100/100/100 for most everything, a few dev boxes only get a slice of FC so its 15/100 or 25/100 and no EFD
donlab1
6 Posts
0
March 24th, 2014 10:00
or symmigrate the extents that live in SATA to FC instead of everything
xubair1
9 Posts
0
March 24th, 2014 13:00
Thank you all who contributed here. After close analysis, we drove to a conclusion that data is indeed suited mostly for sata tier. As a consequence, there is very little that can be done here.
Thanks again for all your inputs.