Solved!
Go to SolutionV
VithalKamesh
4 Posts
0
1455
September 1st, 2011 13:00
Aggregate values on symmetrix performance?
Hi All
Some one can provide some aggregate values on symmetrix performance to verify the utilization,Response Time and Queue Depth. I know the exact values on Clarrion as we normally configure on Navi Analyzer with Utilization - 70 % , Response Time 20 MS and Queue Depth as 10.
Some one can provide the values smillarly on Symmetrix.
Require Thresh hold values on Cache Hit Ratio smillar to Hight water marks and low water marks. Is there any performance document on Symmetrix smillar to this sort of document : h5773-clariion-best-practices-performance-availability-wp.pdf
What is the Cache Hit Ratio values
Is there any parameters Like Low Water mark and High Watermarks on Symmetrix
If the Cache Hit Ration is 100 & is that good as per the symmetrix performance Analyzer report.
Is there any standard performance file where normally EMC follows and send to the customer on symmetrix.
can you share one file as a reference to me. I have my clariion stuff i.e PDF where normally sends by EMC to customer.
Regards
Vithal
VithalKamesh
4 Posts
819
0
Posted September 1st, 2011 21:00
Can you refer any performance document on Symmetrix. Do you have any sample Template document symillar to CLARiiON where EMC normally refers to customer.
VithalKamesh
4 Posts
819
0
Posted September 1st, 2011 21:00
So Is there any High Water Mark level in Symmetrix.
In Clarrion we normally refer to 60-80 water Mark level but in Symmetrix/V-MAX we refer to 70-90 is that correct.
When we pull the V-MAX performance Report we see the Write cache Hit Ratio is @ 100 % is these values or correct.
VithalKamesh
4 Posts
819
0
Posted September 1st, 2011 21:00
Can you come up with accurate values please. This information is not sufficient for me.
Can you refer any document if you have.
So use the same values you do on the CX, 70% utilization, 20ms response times, etc. These are not specific to any storage platform, but are generic to all storage systems. Response time will go up exponentially when you approach 100% utilization, which is why response time is an important metric.
Cache can be tricky in the Symmetrix because it is so large. The main thing to watch out for is high write pending counts. This is not usually a sign of too little cache, but more often too slow a destage for the incoming write rate.
There is a tool that the Symm SPEED gurus have access to that can tell you if adding cache may increase your read hit ratios. There is no real way to look at any of the normal stats and figure this out. Many people use Fall Through Time (FTT) to gauge if the system has too little cache. Normally above 300 seconds is good, but again, it is going to depend on your application. Some applications may have a normal FTT of 150 seconds while others have 600 with the same IO rate and the same amount of cache.
For writes, up to 80% of user cache (75% in 5875 code) can be write pending in the system. 5% of that can be write pending for any one device.
When we go over 50% of the writable cache being write pending, the priority of destage becomes more important. Below that level it is much lower than reads. Above that level it is treated the same as reads, so it can impact response time when above 50%.
That would be the only high water marks that I'm aware of.
AJK3
2 Intern
•
108 Posts
819
0
Posted September 2nd, 2011 01:00
Symmetrix performance values or standard matrix are not as straigh forward as CX. You have various factors involved due to the way the symmetrix handles the IO although the components are similar to CX.
Although if you search for performance documents, you wont be able to find any numbers as they vary according to configuration, host environment, connectivty and several other factors in symmetrix.
As dynamox mentioned, any thing above 50% can be considered as the trouble making one in the future. What tool are you using for Symmetrix performance monitoring.
There are few things that I can refer for initial checking.
Cache Hit Ratio should be 90% for ideal cases and FTT of DAs should not be less than 1000 for longer periods.
All FA utilization should be less than 70% and balanced
Write pending should not be above 50%
Device write pendings should not exceed threshold limit
verify the disk utilization and see if there are any disks that are much busier than the other.
you have lot of disks, DA, FA components involved so it will not easy determine the values right away. Probably dynamox or RRR or Marc or Stefano can or add or change. These are not accurate, just some baselines.
give me a call if you required..