Unsolved
This post is more than 5 years old
1 Rookie
•
61 Posts
0
792
September 24th, 2013 09:00
Increasing spindles correlated to low response times?
Hi,
I have a 18 GB(maximum) of cache on each SP of VNX5700. After the recent code upgrade to R32 P206. SP is utilizing more cache and lowered our read and write cache size.
As a result cache remains utilized over and above 75% throughout the day and response times and queue length have increased. Not to forget we also have FAST cache on this box of about 733 GB, which remains utilized below 65% round the clock. Also, we have Auto-tiering enabled on pools, with SSDs.
I am planning to add more spindles to our storage pools in order to distribute the IO load and lower the queue length and response times.
Am I heading in right direction?
Any suggestions are appreciated.
Regards,
Nishant Kohli
Vipin VK
2 Intern
•
812 Posts
0
September 24th, 2013 23:00
You have to ensure you are following the best practices while adding disks ( Number and type of disks getting added based on the number and type of disks used in pool and RAID types configured). Refer performance and availability best practices guide from https://support.emc.com .
SushantGulati
55 Posts
0
September 27th, 2013 01:00
Hi Nishant,
Welcome to the community.
Can you please specify how much Read Cache and Write Cache sizes do you have configured individually? Also, can you furnish some details regarding the type of cache(RC/WC) which is being utilized continously around 75%? Although the increase in number of spindles is almost always helpful, but this situation needs to be analyzed a little further before you move ahead. Please do make sure that the guidelines specified in the Performance and Availability Best Practices document mentioned by Vipin earlier, are followed.
kelleg
4.5K Posts
0
September 30th, 2013 14:00
On the Write cache, the important metric you want to look at is SP % Dirty Pages. As long as that is staying between the High Watermark and the Low Watermark, the Write cache is performing as designed. If SP % Dirty Pages is constantly at 100%, then you need to add more spindles to the Pool where the Write IO is the heaviest.
glen