Monitoring Symmetrix Performance using symstat CLI
Monitoring Symmetrix Performance using symstatCLI
Introduction
This article describes how to use the Solutions Enabler command symstatto view the performance statistics of Symmetrix Array. The symstatcommand captures, in real-time, statistics information about a Symmetrix array.
Detailed Information
The symstat command performs the following features:
Retrieves the performance counts for the Symmetrix array as a whole
Retrieves the performance counts for a Director or Director Ports
Retrieves the performance counts for Symmetrix devices
Retrieves the performance counts for Device Group, Composite Group, RDF Group
The symstat command provides different types of performance display options:
REQUESTS - Reports I/O requests and throughput for selected devices, directors, or SRDF/A sessions. (This is the default type; if no type is specified, REQUESTS is used.)
BACKEND - Reports back-end I/O requests and throughput for selected devices.
PORT - Reports performance statistics for a director port.
ISCSI - Reports GigE network statistics.
CACHE - Reports cache activity for selected front-end or remote link directors, or SRDF/A sessions
MEMIO - Reports cache memory to disk activity for selected devices.
CYCLE - Reports cycle summary information for SRDF-A sessions
DISK - Reports back-end I/O requests and throughput for selected disks.
PREFETCH - Reports track prefetch disk activity for selected back-end directors only.
RDF - Reports SRDF statistics from the perspective of RA groups, devices, or directors.
DMSP - Reports dynamic mirroring service policy (DMSP) statistics for the selected device(s).
Command Examples:
1. Array Performance Statistics. I/O and throughput statistics can be returned for all devices on a specific array by specifying the Symmetrix ID by -sid, specifying the interval by –iand the number of samples (counts) by –C. ( Symmetrix ID = 250, counts = 100, interval = 60 in following example)
symstat-type REQUEST -sid 250 -i60 -c 100
In command output, IO/Sec READ/WRITE are Read/Write IOPS, KB/Sec READ/WRITE are Throughput per second.
2. Device Performance Statistics. symstat can display performance statistics such as I/O requests via front-end adapters and throughput on devices. For example, for sample intervals of 120 seconds and a sample count of 3 on logical device DEV001 in device group prod_r1, use command:
symstat -i 250 -c 3 -g prod_r1 -ld DEV001
3. Device BACKEND Statistics. Using symstat with the BACKEND type specified,you can display performance statistics such as I/O requests using a DA (back-end director) and throughput on a device.
In command output, IO/sec is IOPS which summarize Host to Front-end directors, disk to Back-end directors, local to Remote link directors. RW is Total read/write cache request rate.
5. Director ports performance statistics. Using symstat with the PORT type specified, you can display performance statistics such as I/O requests (IO/Sec) and throughput (Kbytes/Sec) of director ports for a specific Symmetrix array. For example, for sample intervals of 60 seconds, and a sample count of 3 (for I/O requests and throughput) on Symmetrix 250 all director ports:
symstat -i 60 -c 5 -type PORT -sid 250 -dir ALL
Specifying single director port, use following command:
6. MEMIO performance statistics. Using symstat, you can display performance statistics, such as cache memory to disk I/O on particular device group. For example, enter
symstat-type MEMIO -sid 250 -i 60 -c 100
In command output, WP Tracks are the count not yet destaged to disk. Prefchd/DestgdTracks/SecareTracks per second pre-fetched from disk to cache upon detection of a sequential read stream and
Tracks per second saved into disks. %Dev WPmax is Write-pending device limit percentage which defined on VMAX system level.
7. Disk performance statistics. Using symstat with the DISK type specified,you can display performance statistics, such as I/O requests and throughput on a physical disk. For example, for sample intervals of 60 seconds and a sample count of 3 on disk 02A:C5, enter:
symstat-type disk -i 60 -c 3 -sid250 -disk 2a,C,5
For all SCSI IDs on DA 02A, interface C, use command:
symstat-type disk -i 60 -c 3 -sid250 -disk 2a,C,ALL
8. SRDF performance statistics. The SRDF type statistics extend the statistical information provided by the SRDF director to external applications providing greater visibility into the performance and behavior of Symmetrix arrays in the field. These statistics can be used to monitor on-going activity and to analyze problematic behavior.
To return statistics for all RA group numbers, specify all, use command:
symstat -sid 250 -i 60 -c 2 -type RDF -rdfg 11
For example, to return SRDF device-level statistics for device 37, use command:
symstat -sid 250 -i 60 -c 2 -type RDF -dev 37
To return SRDF director level statistics, use command:
symstat -sid 250 -i 60 -c 2 -type RDF -dir 1d
To return link-level details, specify the -rdflink option as shown in the example below command.
With regards to RA port CPU I/O settings: "Synchronous I/O Percent", "Asynchronous I/O Percent", and "Copy I/O Percent".
What should they be set to in a SRDF/A environment? Does setting the "Asynchronous I/O Percent" to 20% as an example,,, limit the RA processors to only devoting 20% of their processing power to Asynchronous I/O ?
If there is just Asynchronous SRDF traffic active, it will get 100% of the CPU. The QOS values are enforced only if the traffic type is present. If 2 of the 3 traffic types are present, CPU is shared as a ratio of the QOS values.
For example, with the default values are 70/20/10; if there is synchronous and adaptive copy traffic only, the CPU will be shared 7:1. That is, 87.5% for synchronous and 12.5% for adaptive copy.
I have few queries on these commands. After getting output, how we are going to interpret it???
What are the certain scenarios when we need to take actions based on the visible stats.. For example, one application showing continuous 20ms/40ms response time is experiencing performance issue & now we need to take some action...
In short, what are the maximum value of few of the parameters of the output result of all the above mentioned commands, so that we can say that Ok.. The disk(s), cache memory, FA/DA/RDF directors/ports or the I/O requests and throughput activity is facing Write/Read/Any Latency or Performance issue & will go ahead & perform certain activities for remediation of the same.
Appreciate a quick & elaborate reply from your end.
pablosanzoni_bn
22 Posts
7799
0
Posted November 11th, 2013 13:00
Hi,
it is a very useful information,
Thanks for sharing it into the group.-
Regards
Pablo