This post is more than 5 years old
108 Posts
0
1406
SRDF/A & Read Performance on Symmetrix with Sybase
We have a SRDF/A setup on a DMX-3 and its for a Sybase reporting server installed on a Solaris Box. Reports are performing very slow while
SRDF/A is in use and they are performing quite well while the particular SRDF/A group is suspended. I could think this can be an issue with insufficient cache. But is there any other relation wrt performance difference?
Quincy561
1.3K Posts
0
July 15th, 2010 06:00
That isn't very significant, and it doesn't sound like RDF was active for very long. Were the R1 and R2 in sync? Maybe the systems were in copy mode rather than full RDF/A mode?
JasonBailey
147 Posts
0
July 15th, 2010 04:00
thats the only thing I can think of as well
you could check read miss rate and LRU fallthru time to verify
JasonBailey
147 Posts
1
July 15th, 2010 04:00
Its a cache metric
Least Recently Used Ring Fallthru Time
So if you are short on cache, you will cycle your cache more frequently so read miss rate may increase
Quincy561
1.3K Posts
1
July 15th, 2010 04:00
LRU fallthru is a measure of how long it takes for slots to go from the top of the LRU, to then get recycled. It does not count time that they stay in cache being recycled. However it can be very useful for a trending or looking for a change in behavior such as this case. Look for it in the DA statistics.
AJK3
108 Posts
0
July 15th, 2010 04:00
Thanks Jason..
What is LRU Fallthru? I could not get that.
AJK3
108 Posts
0
July 15th, 2010 05:00
RDF/A was not on only for 10-15 mins and the difference is around 30, with on its around 300 and its around 330 with RDF off.
Quincy561
1.3K Posts
0
July 15th, 2010 05:00
The question is what is the difference in the FTT between running with RDF/A on and off?
AJK3
108 Posts
0
July 15th, 2010 05:00
Thanks Jason and Quincy.
I could see that difference and Average LRU time increases during the off bussiness hours and weekends and its very low during the bussiness hours.
And its very large on other arrays where there is no cache contention.
AJK3
108 Posts
0
July 15th, 2010 06:00
RDF/A is continously on. It was suspended just for 15 mins to test the performance. Once link was suspended, performance was very good.
Booyah2
184 Posts
1
July 15th, 2010 08:00
Are you running your report job from the R1 array or maybe performing some sort of Snap/clone at the R2 array?
AJK3
108 Posts
0
July 19th, 2010 03:00
Report is being run at R1 array and there is %hit of 50-70% on it, on the R2 side its in the range of 90+