how many read/write iops do you see to those LUNs ..as well as other LUNs. Did you look in analyzer to see how many IOPS your SAS/NL-SAS are handling ? High response times throughout the day or specific times. What kind of response times do you see ?

Are you using Thin LUN? Thin LUN is not designed for performance. You can try to migrate SQL DB files to Thick LUN or RAID group LUN (traditional LUN). RAID 1/0 would be better.

Anyway, you can collect Analyzer logs (.nar/.naz files) and send to EMC support for analysis.

How to gather the necessary information for a CLARiiON or VNX performance Analysis ?

LUN Utilization: The LUN Utilization metric of the LUN Performance Bottleneck indicator measures/displays how often the LUN is serving an incoming request. A high LUN utilization does not necessarily mean that the LUN is nearing capacity. A LUN is considered busy if any of its disks are busy, so LUN utilization usually presents a "conservative" view.  That is, a high LUN utilization value does not necessarily indicate that the LUN is approaching its maximum capacity, just that at least one of its disks are busy.

Values in the columns are calculated with the folowing:

"% Utilization (avg. over period)"

"% Utilization (current)"

So it is firstly necessary to figure out which disks in the problemetic LUNs were that busy, which types of file were placed in those disks and during which period the LUN utilization ran to 100%.

Suggest you start the analysis with both storage side and application side.

On the storage side, follow this EMC KB:https://support.emc.com/kb/9132 to gather the storage performance report, check what target disk type they are and pay attention to the iops on disk for OLTP and bandwidth on disk for data warehouse in the performance report. Check how the read cache set and if the FAST cache was enabled on the LUNs and pay attention to the cache hit ratio in the performance report.

On the application side, find out the file type in those disks and what kind of database operation was processing in the peak time. Since the utilization indicates disk busy, it is also better to tune the database SQL statements and file logical distribution.

You may want to read over the below doc to get a clearer understanding on SQL and VNX performance.

https://support.emc.com/docu50157_White-Paper:-Microsoft-SQL-Server-Best-Practices-and-Design-Guidelines-for-EMC-Storage-EMC-VNX-Family,-EMC-Symmetrix-VMAX-Systems,-EMC-Xtrem-Server-Products.pdf?language=en_US

also the 10k SAS drives are doing around avg 300 - 350 IOPS and NL SAS drives are barely doing any IOPS I see around less than 30 IOPS on NL SAS drives.

Hi Dynamox,

I am seeing the total iops on the LUN around 5000 iops in which 88 % is READ and the response times around 50ms.

Utilization is pegged at 100 % almost all day.

you are pushing those SAS drives really hard, you need to start thinking about adding more drives to the pool (i would get at least 10 or  15 SAS drives)

Thanks Dynamox

if you have FAST license and running on flare 32, shouldn't it do in-tier FAST so you don't end up with graphs like yours ?

FAST does not move slices based on drive capacity, it moves based on slice performance statistics.

Agreed ..though doesn't look like fast it-tier is doing what it's supposed to do.

Sent from my Verizon Wireless 4G LTE smartphone