SCF Reference Manual for the Storage Subsystem (G06.28+, H06.05+, J06.03+)

The number of cache blocks containing a valid disk block in memory.Blocks In Use
The percentage of allocated blocks that are currently dirty (blocks in cache that have
been changed but are not yet written back to disk).
Blocks Dirty
The percentage of cache calls that the disk process made for user read requests. (When
added together, the percentages displayed for Cache Reads and Cache Writes equal
100 percent.)
Cache Reads
The percentage of cache reads when the requested block was found in cache memory.
(When added together, the percentages displayed for Cache Read Hits and Cache Read
Read Hits
Misses equal 100 percent.) This value should be a high percentage, indicating that the
requested blocks are frequently found in cache. If the percentage is low and you want
to raise it, increase the size of cache by using the ALTER DISK, CACHE command. For
more information, see “Configuring the Size of Disk Cache” (page 86).
The percentage of cache reads when the disk process could not find the requested block
in cache and had to bring the block in from disk. A user write request does not affect
Read Misses
the read counters even if the cache write request causes a disk read. If this percentage
becomes very high, consider increasing the size of cache. For more information, see
“Configuring the Size of Disk Cache” (page 86).
The percentage of cache calls that the disk process made for user write requests and
that resulted in cache writes. (When added together, the percentages displayed for
Cache Reads and Cache Writes equal 100 percent.)
Cache Writes
The percentage of cache writes for which the block was found changed (dirty). When
new data is inserted into a block and the disk process finds a dirty block, it does not
Write Dirties
have to perform a disk read but it must perform a disk write. Having a large value for
Cache Write Dirties reduces the number of required physical I/O operations (that is,
disk reads and disk writes).
The percentage of cache writes for which the block was found unchanged in cache (not
dirty). When new data is inserted into a block, the disk process does not have to perform
Write Cleans
a disk read but it must perform a disk write. This counter is also incremented when an
end of file (EOF) is extended (an application appends to the end of a file).
The percentage of cache writes that resulted in the disk process not finding a block and
having to read the block from the disk. For cache write misses, the disk process must
Write Misses
perform both a disk read and a disk write. (When added together, the percentages
displayed for Cache Write Cleans, Cache Write Misses, and Cache Write Dirties equal
100 percent.)
The total number of cache reads and cache writes performed during the measurement
interval. If a user request bypasses cache (for example, by specifying direct I/O), none
of the cache counters are affected by that call.
Cache Calls
The number of times a cache call expected to find the block in cache but could not, due
to limited memory space. This number should be very small or zero. If it is not, there is
Cache Faults
insufficient memory to satisfy the cache configuration. You must either add more memory
or reconfigure cache to use memory more efficiently.
The number of times an audited dirty block required an audit-trail write to disk to make
room for a new block to be read from disk. (This value is the same as the
AUDIT-BUF-FORCE counter in the disk report from the Measure product.)
Audit Forces
Reconfiguring Cache to Resolve Performance Problems
Based on the statistics displayed by a STATS DISK report, you can reconfigure cache to resolve
memory performance problems
What to DoCondition
Use the ALTER DISK, CACHE command to increase the number of
cache blocks for that cache block size.
Cache Read Hits is low. Cache Read Misses is
high.
Use the ALTER DISK, CACHE command to reduce the number of
cache blocks for that cache block size.
Cache Faults is large.
Displaying Disk Cache Statistics 61