Solved!
Go to SolutionCX4 Perfomance
Hello all,
After reading the document "EMC Clariion Best Practices for Performance and Availability" and merge 7 days worth of Analyzer data, I have question that I would like to ask you guys for help.
How can I identify the I/O type on a specific LUN, RAID Group and/or Storage Pool?
In other words,
How can I know if the data read/written to my array is random or sequential?, the block size?
Thank you very much!
you can look at average seek distance (for the disks) to see how small the distance is in GB, the smaller the number the more sequential the IO .
if the IO are wtries , you can look at total WR throughput and full stripe Write(for LUN), the closer these numbers are together the more sequential the IO.
for reads looks ath total read througput and RD cache hits. RD cache hits 100% means 100% sequential reads. if the Read IO are purely random, the RCHR(read cache hit ratio) can be as low as 5%. Also multi-stream multi threaded sequential acess can also lowers the RCHR.
You can also look at Support Solution emc218359 - this is a master article that lists all the solutions for using Analyzer and performance issues.
You can also go to Monitoring/Analyzer/Customize and on the General tab, select Advanced to see more counters in Analyzer when you view a archive file - the lower left panel
glen
Also, please remember to mark your question "Answered" when you get the correct answer.
SteveZhou
2 Intern
•
136 Posts
355
0
Posted December 27th, 2011 18:00
you may select a specific object that you want to check, such as LUN, disk, SP, RAID Group, etc. Then to choose any counters you want, like read size, write size, throughput, Utilization, etc.
To determine the I/O patten, you should check lun object, choose read or write size to determine the I/O size. But for random or sequential, there is no such a counter could tell us directly. You need to check counters like "SP Cache Read/Write Hits/Ratio", "Disk Crossings(%)", etc, but again, it won't be easy to determine. Basically, you should understand your application behaviour to have a general view of the possible I/O patten. Like OLTP is usually for "small, random 70% read with 30% Write".
-Steve