2010-09-18 08:03 AM
I have a fas 2020 filled with 15k SAS drives.
those drives are one one controller.
I also have shelf attached with 6 SATA 7200 drives
which are on another controller.
On this drives I have 2 volumes, cifs and another
for vmware. Agregate is Raid DP with 0 spare disks.
If I run Crystal Disk Mark on virtual machine on sata drives
I get 50 MB. Servers also seem slow, if you open a program it takes a while, sometimes
even if you go to start - programs - accesoris it takes 4 seconds to show items inside....
But If I move this virtual machine to SAS drives, it's ok.
I also moved that same machine on qnap storage with 5 SATA drives and I
got much better result in Crystal Disk Mark - 89 MB the same virtual machine.
I have expected to be on pair with qnap but not that slow.
What can be done to improve speed of SATA aggregate?
2010-09-20 01:39 AM
analysis of post..... done
possible bottlenecks in this setup...listed
a) filer side I/O
b) network side Througput
c) host system memory & cpu management
d) VMware host setup
e) VMware guest setup
sorting in order if likehood based on personal experience...skipped...
begin of human interaction...
let us assume the netapp is the problem, i would start looking at the raid group design
if i had found out about netapp-sided disk i/o shortcomings.
(one way to find out is to open a support case, and run the 'perfstat' analysis tool over the installation)
if you have a large raidgroup on which the aggregate, and subsequently volumes resides - then you have
a better disk i/o in comparision to having several small raidgroups - due to the combination of spindles and r/w heads.
This is true only to a certain extend (imagine a logarytmic curve).
However, direct access of data (you refer to your laptops sata drive) may still be faster than any kind of access through a network
no matter how expensive your storage backend is. Despite Network bandwidht limitations, you get also the protocol overhead
from SMB (most overhead) or NFS (least overhead).
("user wants file, is he allowed to? dont ask me, ask DC. Read, write or change? dunno, lemme ask DC..... DC says change... check.. ok lets give him that file.... check")
The good news is, your data is protected by industry proven raid technology (unfortunately you have 0 spares, which kind of takes this feature away,
after a disk fail your volume will be "degraded" 24hrs later in order to protect your data - it goes to read only. Yes, its tweakable, no i wouldnt do it.
A more spohisticated answer is difficult, the beauty of a filer is many can use it at same time, and still get a decent speed.
try having 500 users access that single laptop-sata drive
Support case may give you certainity whether there is anything tweakable on the netapp.
If you run ESX server, then there would be a number of things to verify for "best practises" as well. But it was
not entirely clear what vmware software youre using.
Probably not the best Answer one can give, but i figured i cannot let your question remain