TPDC Manual (G06.24+)

Using the SPAM Performance Data Charts and
Reports
Tandem Performance Data Collector (TPDC) Manual—526357-001
5-62
Disc Subsystem Activity (DSCREPT)
DISC SUBSYSTEM Activity Report for System \CSO (Release D20)
------------------------------
$Volume LDV RQ:sec IQlen IQmsec Blkd % (headings continued below)
======== === ------ ------ ------ ------
(1) (2) (3) (4) (5) (6)
<--- CACHE:RQ ----> <--------- BUSY % --------> <--- I/Os PER REQUEST ---->
Hit Miss Flt ABF PPID PCU Total Read Write Seek Total Read Write Seek
---- ---- ---- ---- ------ ---- ------ ------ ------ ------ ------ ------ ------ ------
(7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20)
THIS REPORT SHOWS DISC SUBSYSTEM ACTIVITY OVER THE MEASUREMENT INTERVAL (WINDOW). FOR THE PURPOSES
OF THIS REPORT A DISC SUBSYSTEM IS COMPOSED OF DISC PROCESSES (INCLUDING DISC CACHE) AND THE
PHYSICAL DISC DEVICE(S) THAT COMPRISE A LOGICAL VOLUME; THE LEFT HALF OF THE REPORT SHOWS DISC
PROCESS AND DISC CACHE ACTIVITY, WHILE THE RIGHT HALF OF THEREPORT SHOWS PHYSICAL DISC ACTIVITY.
(1) The disc volume name. (7)-(10) (cont.)
(2) The Logical Device Number of the disc volume. (9) The average number of times per request that
cache tables indicated that a needed block was
(3) The average number of requests for service in cache, but the Memory Manager had 'stolen'
received per second. the underlying memory for another purpose,
aggregated over all cache block sizes. This
(4) The average number of requests for service typically indicates either too large a cache
on the the Disc Processes' 'internal' queue. size, or too little real memory for the CPU.
(5) The average time in milliseconds that (10) The average number of times per request that
requests for service spent in the Disc the Disc Process had to write to the TMF audit
Processes’ 'internal' queue. trail volume before it could write a 'dirty'
cache block and so free that block for another
(6) The percentage of requests for service that use, aggregated over all cache block sizes.
encountered record or file locks (and were This value should be close to zero.
therefore impeded from further forward
progress until locks were released). This (11) The <cpu>,<pin> of the primary Disc Process.
value tends to show a strong positive
correlation to both (4) and (5), above. (12) The Controller # and Unit # of the disc(s)
under control of the primary Disc Process.
(7) A group of columns showing the key Disc
thru Cache activities per request for service (13) The percentage of the window that the disc
(10) received by the Disc Processes: thru was busy, in total (13) and in each activity:
(16) reading (14), writing (15) and seeking (16).
(7) The average number of times per request
that a required block was found in cache (17) thru (20)
on a read or buffered write operation,
aggregated over all cache block sizes; The average number of I/O operations per
each ‘hit’ saves a disc I/O. request for service, in total (17) and in each
activity: reads (18), writes (19) and seeks
(8) The average number of times per request that (20).
a required block was not found in cache on
a read operation, aggregated over all cache
block sizes; each 'miss' causes a disc I/O.
Not all disc requests go through the Disc For mirrored volumes, there is a line in the
Cache (for example, writes to TMF audit right half of the report for each of the physical
trails bypass the cache). discs that comprise the volume; zero values are
suppressed.