performance stats

Question

Hi,

Can some one give the basic under standing of these perf stats and their idle values.

1.Read response time

2.Write response time

3.Hit rate

4.Miss rate

If possible pls post me a good DOC.

Thanks In Advance

dynamox · Accepted Answer

The following table lists and defines metrics for Symmetrix devices.

% hit	100 * (total hits per sec / total ios per sec) Percentage of I/O operations, performed by the Symmetrix device, that were immediately satisfied by cache.
% miss	100 - (% hit) Percentage of read and write miss operations performed each second by the Symmetrix device that were misses. A miss occurs when the requested read data is not found in cache or the write operation had to wait while data was destaged from cache to the disks.
% random read hit	100 * (read hits per sec / total ios per sec)
% random read miss	100 * (read misses per sec / total ios per sec)
% read	100 * (total reads per sec / total ios per sec) Percentage of read I/O operations performed by the Symmetrix device.
% read hit	100 * (total read hits per sec / total reads per sec) Percentage of read hit I/O operations performed each second by the Symmetrix device. Read hits occur when the read operation is satisfied directly from cache.
% read miss	100 * (total read misses per sec / total reads per sec) Percentage of read miss I/O operations performed by the Symmetrix device. Read misses occur when the data requested by the read I/O operation was not in cache, causing a wait while it was fetched from disk.
% sequential io	100 * (seq ios per sec / total ios per sec) Use this metric for mainframe performance analysis only.
% Sequential read	100 * (seq reads per sec / total ios per sec) Use this metric for mainframe performance analysis only.
% sequential writes	100 * (seq writes per sec / total ios per sec) Use this metric for mainframe performance analysis only.
% write	100 * (writes per sec / total ios per sec) Percentage of total write I/O operations performed each second by the Symmetrix device.
% write hits	100 * (write hits per sec / writes per sec) Percentage of write hit I/O operations performed by the Symmetrix device. A write hit occurs when the I/O write operation is received, it is immediately staged in cache and will be written to disk at a later time.
% write miss	100 * (write misses per sec / writes per sec) Percentage of write I/O operations performed each second by the Symmetrix device that were write misses. A write miss occurs when the I/O write operations are delayed because the system or device write-pending limit was reached and the cache had to destage slots to the disks before the writes could be written to cache.
average io size in Kbytes	(HA Kbytes transferred per sec) / (total ios per sec) Average size of an I/O operation performed by the Symmetrix device.
average read size in Kbytes	(Kbytes read per sec) / total reads per sec) Average size of a read I/O operation performed by the Symmetrix device.
average write size in Kbytes	(Kbytes written per sec) / (writes per sec)
DA Kbytes read per sec	Number of kilobytes read by the disk directors from the disk each second.
DA Kbytes transferred per sec	(DA Kbytes read per sec) + (DA Kbytes written per sec) Sum of the kilobytes read and written each second on all disk directors associated with the Symmetrix device.
DA Kbytes written per sec	Number of kilobytes written to the disk from the disk director each second.
DA prefetched tracks per sec	Total prefetched tracks each second from the disk directors to the cache.
DA prefetched tracks Used per sec	Number of prefetched tracks used each second from the disk directors to the cache.
DA read requests per sec	Number of read requests each second performed by the disk directors to the cache.
DA write requests per sec	Number of write requests each second performed by the disk directors to the cache.
HA Kbytes transferred per sec	(Kbytes read per sec) + (Kbytes written per sec) Total number of kilobytes read and written each second through host directors associated with the Symmetrix device.
hits per sec	Number of I/O operations performed each second by the Symmetrix device that were immediately satisfied by cache.
ios per sec	Number of I/O operations performed each second by the Symmetrix device, including writes and random reads. In contrast, the total ios per secmetric includes writes, random reads, and sequential reads.
Kbytes read per sec	Number of kilobytes read by the Symmetrix device each second.
Kbytes written per sec	Number of kilobytes written by the Symmetrix device each second.
max write pending Threshold	Maximum number of write-pending slots available for the Symmetrix device. Max Write Pending Threshold is not a static number. It depends on Symmetrix activity. Each Symmetrix device is assigned a limit of write-pending slots that can dynamically change between a base value and a value three times the base (the maximum value). Once the Max Write Pending Threshold has reached three times the base value, writes to the device are delayed so that the cache can destage, which frees the cache slots. As cache slots are freed, the writes resume. While the write-pending limit is reached, disk directors operate in a priority destage write mode. This gives write data higher priority than usual. During the delay, writes to this Symmetrix device are counted as write misses. Note: If you have a newer DMX, such as a DMX3 or DMX4, the device limit is 5% of the system write pending limit.
read hits per sec	Number of random read hit I/O operations performed each second by the Symmetrix device. Read hits occur when the read I/O operation is satisfied directly from cache. The read hits per sec metric for the Symmetrix device statistic does not include sequential read hits. In contrast, the metric total read hits per secincludes random and sequential read hits per sec.
read misses per sec	(reads per sec) - (read hits per sec) Read misses occur when the data requested by the read I/O operation was not in cache, causing a wait while it was fetched from disk.
random reads per sec	Number of I/O operations performed each second by the Symmetrix device that were random reads. This Random Reads per sec metric for the Symmetrix device statistic does not include sequential reads. In contrast, the metric total read hits per secincludes random and sequential read hits per sec.
sampled average reads missed time (ms)	Completion time of a read-miss as measured by the host director. Measurements are taken for a sample set of approximately 30% of the I/Os. Read misses occur when the data requested by the read I/O operation was not in cache, causing a wait while it was fetched from disk. Metadevices: This is approximately (within milliseconds) the same response time that the host sees at the head of the meta. The value is calculated by dividing the total sampled elapsed time by the number of sampled events. The number of sampled events counter, and sampled elapsed time counter are maintained for all metamembers under the statistics for the head of the meta.
sampled average read time (ms)	Completion time of a read as measured by the host director. Measurements are taken for a sample set of approximately 30% of the I/Os. Metadevices: This is approximately (within milliseconds) the same response time that the host sees at the head of the meta. The value is calculated by dividing the total sampled elapsed time by the number of sampled events. The number of sampled events counter, and sampled elapsed time counter are maintained for all metamembers under the statistics for the head of the meta.
sampled average write time (ms)	Completion time of a write as measured by the host director. Measurements are taken for a sample set of approximately 30% of the I/Os. metadevices: This is approximately (within milliseconds) the same response time that the host sees at the head of the meta. The value is calculated by dividing the total sampled elapsed time by the number of sampled events. The number of sampled events counter, and sampled elapsed time counter are maintained for all metamembers under the statistics for the head of the meta.
sampled average WP disconnect time (ms)	Completion time of a write-miss as measured by the host director. Measurements are taken for a sample set of approximately 30% of the I/Os. A write miss occurs when a write I/O is held while data is destaged from cache. Metadevices: This is approximately (within milliseconds) the same response time that the host sees at the head of the meta. The value is calculated by dividing the total sampled elapsed time by the number of sampled events. The number of sampled events counter, and sampled elapsed time counter are maintained for all Metamembers under the statistics for the head of the meta.
sampled RDF writes waits per second	Number of I/O operations performed while waiting for an RDF transfer. This is simply a counter.
seq ios per sec	(seq reads per sec) + (seq writes per sec) Use this metric for mainframe performance analysis only.
seq read hits per sec	Portion of the explicitly counted sequential reads that were satisfied from the cache on request. Information about the type of environment you are working in. Mainframe environment: The Sequential read hits per sec is useful as a sequential counter. Open System environment: The SCSI I/O command does not have indicators to identify whether the I/O is part of a sequential stream, so the counts are not actual indicators of full sequential activity. This metric is needed for open systems to calculate the metric total reads per sec. This metric must be added to the read hits per sec to get the value for the Total read hits per sec.
seq reads per sec	Information about the type of environment you are working in. Mainframe environment: The seq reads per sec is useful as a sequential counter. Open System environment: The SCSI I/O command does not have indicators to identify whether the I/O is part of a sequential stream, so the counts are not actual indicators of full sequential activity. In an open-system environment the Sequential Read metric indicates to the back end to start pre-fetch activity after two sequential read misses occurred in a row. In the case where the data is in cache and all reads are hits, the back end does not get involved and therefore there is no indication that the read is sequential. This metric is still needed for open systems to calculate the total reads per sec metric. Since sequential read misses are not included in the read counters the seq reads per sec must be added to the Random Reads per sec to get the value for the Total Reads per sec.
seq writes per sec	Number of write I/O operations performed each second that were sequential. Use this metric for mainframe performance analysis only.
system bus Kbytes per sec	(HA Kbytes transferred per sec) + (DA Kbytes transferred per sec)
total DA req per sec	(DA read requests per sec) + (DA write requests per sec)
total hits per sec	(total read hits per sec) + (write hits per sec) Total number of I/O operations performed each second by the Symmetrix device that were immediately satisfied by cache.
total ios per sec	(total reads per sec) + (writes per sec) Total number of read I/O and write I/O operations performed each second by the Symmetrix device.
total misses per sec	(total ios per sec) - (total hits per sec) Total number of missed I/O operations performed each second by the Symmetrix device.
total read hits per sec	(read hits per sec) + (seq read hits per sec) Total number of read hits and sequential read hit operations performed each second by the Symmetrix device.
total reads per sec	(random reads per sec) + (seq reads per sec) Total number of random read and sequential read operations performed each second by the Symmetrix device.
write hits per sec	Number of write hit operations performed each second by the Symmetrix device. A write hit occurs when the I/O write operation is received, it is immediately staged in cache and will be written to disk at a later time.
write misses per sec	(writes per sec) - (write hits per sec) This number of write misses that occurred for the Symmetrix device each second. A write miss occurs when the I/O write operations are delayed.
write pending count	Number of cache slots that were write pending for the logical volume at a point in time. This number changes according to the cache destage activity rate and the number of writes. A write is pending when it has been written to cache but has not yet been written to the disk.
writes per sec	Number of write I/O operations performed each second by the Symmetrix device.

Symmetrix disk metrics

The following table lists and defines metrics for Symmetrix disks.

% busy	100 - %disk idle
% idle	Disk Idle Time Per Sec * 100
% utilization	This metric shows how close the disk performance is to 100% utilization. Percentage is based on the number of SCSI commands being performed by the disk.
accumulated queue depth	Sum of queue depths for the disk when an event enters this range (0 - over 640).
average hypers per seek	(seek distance per sec) / (seeks per sec) Average number of hypervolumes that the disk head crossed during one seek.
average Kbytes per read	(Kbytes read per sec) / (read commands per sec)
average Kbytes per write	(Kbytes written per sec) / (write commands per sec)
average busy queue depth	Average Queue Depth / (1 - Disk Idle Time Per Sec) Ratio of Average Queue Depth to the time the disk is busy.
average queue depth	Calculated against the Accumulated Queue Depth and Total SCSI Commands Per Sec.
average read time	(total read time per sec) / (read commands per sec)
average write time	(total write time per sec) / (write commands per sec)
disk idle time per sec	Average time per second that the disk is idle.
Kbytes read per sec	Number of kilobytes read by the Symmetrix disk each second.
Kbytes written per sec	Number of kilobytes written by the Symmetrix disk each second.
read commands per sec	Number of read operations performed each second by the Symmetrix disk.
seek distance per sec	Number of hypervolumes that the disk head crossed (during all seeks) each second.
seeks per sec	Number of times each second that the disk head moved to find data.
skip mask commands per sec	Number of skip mask commands performed each second by the disk. Skip Mask enables a fragmented track, which is chunks of data with gaps in between, to be read from disk or written to disk in one I/O operation. Without Skip Mask, each chunk of data is read or written in a separate I/O operation, causing a fragmented track to require multiple I/O operations to read or write.
total SCSI commands per sec	Total number of read commands, write commands, skip mask commands, verify commands, XOR write commands, and XOR write-read commands performed by the Symmetrix disk each second.
verify commands per sec	Number of verify commands performed each second by the disk. Verify commands are background scrubbing operations performed periodically to check data integrity on the drives.
write commands per sec	Number of write operations performed each second by the Symmetrix disk.
XOR write commands per sec	Number of exclusive OR (XOR) write commands performed each second by the disk. XOR commands are used to establish parity protection in RAID-S and RAID 5 configurations. In a RAID-S or RAID 5 environment, 3 physical I/Os for each write operation are generated: write, XOR-write-read, and XOR-write. When a write is issued to a disk, which is a member of a RAID-S or RAID 5 set, the write is immediately followed by an XOR-write-read command. The XOR-write-read command extracts the necessary information for parity protection. The information is then written to the parity disk using the XOR-write command.
XOR write-read commands per sec	Number of exclusive OR (XOR) write-read commands performed each second by the disk. XOR commands are used to establish parity protection in RAID-S and RAID 5 configurations. In a RAID-S or RAID 5 environment, 3 physical I/Os for each write operation are generated: write, XOR-write-read, and XOR-write. When a write is issued to a disk, which is a member of a RAID-S or RAID 5 set, the write is immediately followed by an XOR-write-read command. The XOR-write-read command extracts the necessary information for parity protection. The information is then written to the parity disk using the XOR-write command.

Symmetrix disk director metrics

The following table lists and defines metrics for disk (back-end, or dadir) directors.

% read	100 * (reads per sec / requests per sec) Percentage of read requests performed by the disk director.
% seq reads	100 * (prefetched tracks per sec / (prefetched tracks per sec + reads per sec)) Percentage of read requests that were sequential. A sequential read occurs when the tracks brought into cache by prefetch tasks are read in the order in which they were put in cache.
% utilization	This metric shows how close the disk director (back-end director) is to 100% performance utilization. Percentage is based on the number of I/O operations being performed by the disk director.
% write	100 * (writes per sec / requests per sec) Percentage of write requests performed each second by the disk director.
accumulated director idle time	Number of ticks that elapsed while the director did not have any requests in its queue.
average lru0 fall through time	Average time it takes a cache slot in LRU0 to be freed up. It is the average time from the first use of the contents to its reuse by another address. This is a key metric that Symmetrix algorithms use to manage the cache. It is a measure, on average, in seconds, of how long it takes a slot to go from the top of the LRU to the bottom of the LRU for reselection. The time only measures from the last time the slot is placed at the top of the LRU to the time it reaches the bottom of the LRU. Each time a slot is placed at the top of the LRU, the time value used to calculate the Fall Through Time is reset to current time (effectively, zero). Average is calculated using between 4000 - 8000 slots. This metric is a sometimes misunderstood Symmetrix statistic. The cache algorithms may have placed the slot at the top of the LRU many times before the slot finally reaches the bottom of the LRU. The total time the slot is in the cache would have to account for all of the times the slot was placed at the top of the LRU. This would equal cache residency time (time of reselection minus the time of entry). A common mistake is to think that Fall Through Time is cache residency time. Cache residency time does not exist in Symmetrix statistics.
ios per sec	Number of I/O operations performed each second between the Symmetrix disk director and the disk.
long misses per sec	Number of times each second that requested data was not in cache, and was not in the process of being fetched.
queue depth count	Number of events that are waiting while the director's queue is in this range (0 - over 640).
permacache requests per sec	Number of requests between permacache and the disk director each second. Permacache is the part of cache reserved for a particular application. A request is the protocol used by directors (host and disk) to marshal data to and from the cache.
prefetched tracks per sec	Number of tracks prefetched to cache each second by the disk director.
reads per sec	Number of read requests performed each second by the Symmetrix disk director.
requests per sec	Number of request operations performed by the disk director. A request is the protocol used by directors (host and disk) to marshal data to and from the cache. This metric is made up of reads and writes.
short misses per sec	Number of times each second that requested data was not in cache, but was being fetched when the request was executed.
tracks not used per sec	Number of prefetched tracks per sec that have not been used to satisfy a read request. For more information, see prefetched tracks per sec.
tracks used per sec	Number of prefetched tracks per sec that were used to satisfy read requests. For more information, see prefetched tracks per sec.
writes per sec	Number of write requests performed each second by the Symmetrix disk director.

Symmetrix host director metrics

The following table lists and defines metrics for ESCON, fibre, Ficon, and SCSI host (front-end, or sadir) directors.

% busy	100 - %idle This metric is available only for Fibre Channel directors.
% hit	100 * (hits per sec / requests per sec) Percentage of requests performed by the host director and immediately satisfied by cache.
% idle	Ratio of Accumulated Director Idle Time to interval time. This metric is available only for Fibre Channel directors.
% read	100 * (reads per sec / requests per sec) Percentage of read requests performed by the host director.
% read hit	100 * (read hits per sec / reads per sec) Percentage of read hit requests performed by the Symmetrix host directors. Read hits occur when the read I/O operation is satisfied directly from cache.
% utilization	This metric shows how close the host director (front-end director including Escon, fibre, or SCSI) is to 100% performance utilization. Percentage is based on the number of I/O operations being performed by the host director.
% write	100 * (writes per sec / requests per sec) Percentage of write requests performed each second by the host director.
accumulated director idle time	Number of ticks that elapsed while the director did not have any requests in its queue.
accumulated queue depth	Sum of queue depths for the director when an event enters this range (0 - over 640). This metric is available only for Fibre Channel directors.
average queue depth	Average queue depth for the director when an event enters this range (0 - over 640). This metric is available only for Fibre Channel directors.
device write pending events	Number of times, each second, that the write-pending limit for a specific Symmetrix device was reached. When the limit is reached, additional write I/O operations are deferred while waiting for data in cache to be destaged to the disk.
hits per sec	Number of read and write requests performed each second by the host director that was immediately satisfied by cache.
ios per sec	Number of I/O operations performed each second by the Symmetrix host director. This metric represents activity between the Symmetrix and the host or SAN device.
queue depth count	Number of events that are waiting while the director's queue is in this range (0 - over 640).
read hits per sec	(requests per sec) - (writes per sec) - (read misses per sec) Number of read hit requests performed each second by the Symmetrix host directors. Read hits occur when the read requests is satisfied directly from cache.
read misses per sec	Number of read miss requests performed each second by Symmetrix host director. Read misses occur when the data requested by the read I/O operation was not in cache, causing a wait while it was fetched from disk.
reads per sec	Number of random read requests performed each second by Symmetrix host director.
requests per sec	Number of cache slots accessed each second by host directors. A request is the protocol used by directors (host and disk) to marshal data to and from the cache. Requests are made up of reads and writes.
slot collisions per sec	Number of slot collisions each second. A slot collision occurs when two or more directors try to access the same cache slot and the slot happens to be locked for an update operation by one of the directors.
system write pending events	Number of times each second that write activity was heavy enough to use up the system limit set for write tracks occupying cache. When the limit is reached, writes are deferred until data in cache is written to disk.
writes per sec	Number of write requests performed each second by the host directors.

Symmetrix port metrics

The following table lists and describes metrics for Symmetrix ports.

% utilization	This metric shows how close the port performance is to 100% utilization. Percentage is the throughput in kilobytes each second being performed by the port divided by a fixed number based on the type of port.
average io size in Kbytes	(throughput in Kbytes per sec) / (ios per sec) Average number of kilobytes transferred through the Symmetrix port per I/O operation.
ios per sec	Number of I/O operations performed each second by the Symmetrix port.
throughput in Kbytes per sec	Number of kilobytes transferred through the Symmetrix port (entering the Symmetrix system) each second.

Symmetrix system metrics

The following table lists and describes metrics for the total of Symmetrix systems or units.

% hit	100 * (hits per sec / total ios per sec) Percentage of I/O operations performed by all of the Symmetrix devices, for which the read data was in cache and the write operation could be sent directly to cache without having to wait for data to be destaged from cache to the disks.
% reads	100 * (total reads per sec / total ios per sec) Percentage of total read I/O operations performed each second by all of the Symmetrix devices.
% sequential io	100 * (seq reads per sec / total ios per sec) Percentage of sequential I/O operations. Sequential I/O operations are a combination of the read and write operations. A sequential read occurs when the tracks brought into cache by prefetch tasks are read in the order in which they were put in cache. A sequential write occurs when the data is written to cache sequentially.
% writes	100 * (writes per sec / total ios per sec) Percentage of total write I/O operations performed by all of the Symmetrix devices.
deferred writes per sec	A deferred write is a write hit. A deferred write occurs when the I/O write operations are staged in cache and will be written to disk at a later time.
delayed dfw per sec	A delayed deferred fast write (DFW) is a write-miss. A delayed DFW occurs when the I/O write operations are delayed because the system or device write-pending limit was reached and the cache had to destage slots to the disks before the writes could be written to cache.
destaged tracks per sec	Number of tracks written per sec from cache to disks.
device max wp limit	Maximum number of write-pending slots for the entire Symmetrix. System write-pending limit is equal to 80% of the available cache slots. Symmetrix write-pending limit is not simply a sum of all Symmetrix device write-pending slots. It depends on other factors such as cache size and the Symmetrix configuration.
hits per sec	Number of I/O operations performed each second by all of the Symmetrix devices for which the read data was in cache and the write operation could be sent directly to cache without having to wait for data to be destaged from cache to the disks.
ios per sec	Number of I/O operations performed each second by all Symmetrix devices including writes and random reads. In contrast, the total ios per secmetric includes writes, random reads, and sequential reads.
Kbytes read per sec	Number of kilobytes read by all of the Symmetrix devices each second.
Kbytes transferred per sec	(Kbytes Written per sec) + (Kbytes Read per sec) Number of kilobytes written and read by all of the Symmetrix devices each second.
Kbytes written per sec	Number of kilobytes written by all of the Symmetrix devices each second.
number format pending tracks	Number of format pending tracks.
number free permacache slots	Number of available reserved cache slots.
number used permacache slots	Number of use reserved cache slots.
number write pending tracks	Number of tracks in cache that are waiting to be destaged to disk and cannot be overwritten.
prefetched tracks per sec	Number of tracks prefetched to cache each second across the Symmetrix system.
read hits per sec	Number of read hit I/O operations performed each second by all of the Symmetrix devices. Read hits occur when the read I/O operation is satisfied directly from cache.
reads per sec	Number of random reads performed each second by all of the Symmetrix devices. The reads per sec metric for the Symmetrix device statistic does not include sequential reads. In contrast, the total reads per secmetric includes the random and sequential reads each second.
seq reads per sec	Number of sequential read I/O operations each second. A sequential read occurs when the tracks brought into cache by prefetch tasks are read in the order in which they were put in cache.
total ios per sec	(Random Reads per sec) + (seq reads per sec) + (writes per sec) Total number of read I/O and write I/O operations performed each second by all of the Symmetrix devices.
total reads per sec	(Random Reads per sec) + (seq reads per sec) Total number of random read and sequential read operations performed each second by all of the Symmetrix devices.
write hits per sec	Number of write hits performed each second by all of the Symmetrix devices. A write hit occurs when the I/O write operation is received, it is immediately staged in cache and will be written to disk at a later time.
writes per sec	Number of total writes performed each second by all of the Symmetrix devices.

emcsanadmin · Answer

Good info....

praveen56 · Answer

Thanks Dynamox and Gabrial.Dynamox you are the man.

praveen56 · Answer

I did.Thanks

SymmetrixForums · Answer

Did the replies to your post answer your question? If so, please mark the helpful and correct posts and also mark the post as answered. Doing so helps others identify the most useful posts and also serves to award points to those who took the time to respond to your question.

SAN_AK · Answer

good thread

Symmetrix

performance stats

Was this post helpful?