HSMS provides an interface to the software monitor openSM2 for improving the migration procedure. The SM2 monitoring system supplies statistics on the performance of the DP system and the utilization of the resources. During a monitoring period, SM2 regularly – say, every minute – records HSMS information and outputs it to a global SM2 file. All this HSMS information can be accessed by SM2 users, who can analyze some general parameters and follow their development. The users can analyze some general parameters and follow their development. For more detailed information on openSM2, refer to the “openSM2” manual [17].
A migration or recall request is monitored on the host on which it is executed, but only if it is started and ended within the same monitoring period. If in a collector request only one request is to be monitored, all the requests in the collector request are monitored nevertheless.
The following HSMS information is passed to SM2:
Total number of mounted magnetic tapes for all migration and recall activities
Total size of all migrated files (in PAM-PAGES)
Number of migration runs
Total number of extents of the migrated files
Number of files migrated from S0 to S1
Number of files migrated from S0 to S2
Number of files migrated from S1 to S2
Total number of days between the last time a file was used and its migration from S0 to S1
Total number of days between the last time a file was used and its migration from S0 to S2
Total number of days between the migration of a file to S1 and its migration to S2
Number of recall runs (explicit recall runs and implicit recall events)
Total number of days a file spent on S1 before it was recalled to S0
Total number of days a file spent on S2 before it was recalled to S0
Number of files that were explicitly recalled from S1
Number of files that were explicitly recalled from S2
Number of files that were implicitly recalled from S1
Number of files that were implicitly recalled from S2
Number of recall requests that have lasted the following length of time (in minutes): less than 2, 2 thru 4, 4 thru 6, 6 thru 8, 8 thru 10, 10 thru 12, 12 thru 14, 14 thru 16, 16 thru 18 and more than 18 minutes
Number of recall requests that have waited the following length of time (in minutes) from generation to processing: less than 2, 2 thru 4, 4 thru 6, 6 thru 8, 8 thru 10, 10 thru 12, 12 thru 14, 14 thru 16, 16 thru 18 and more than 18 minutes
Number of recall requests that have waited the following length of time (in minutes) from the start of their processing until the ARCHIVE call: less than 2, 2 thru 4, 4 thru 6, 6 thru 8, 8 thru 10, 10 thru 12, 12 thru 14, 14 thru 16, 16 thru 18 and more than 18 minutes
Number of ARCHIVE processing operations for recall requests that have lasted the following length of time (in minutes): less than 2, 2 thru 4, 4 thru 6, 6 thru 8, 8 thru 10, 10 thru 12, 12 thru 14, 14 thru 16, 16 thru 18 and more than 18 minutes
Number of magnetic tape cartridges for recall requests that were mounted for the following length of time (in minutes): less than 2, 2 thru 4, 4 thru 6, 6 thru 8, 8 thru 10, 10 thru 12, 12 thru 14, 14 thru 16, 16 thru 18 and more than 18 minutes
With SM2 and the collected information, the HSMS administrator can adjust the migration parameters appropriately, as the following examples show.
Example 1
Finding:
The number of files per migration run is low.Reason:
The period of time between two migration runs is probably too short. The HSMS administrator is starting some unnecessary migration runs.
Example 2
Finding:
The average time before migration of a file from S0 is short.Reason:
The amount of usable storage space on the disk is too small for the number of users. Either there are too many users or there are too many files which cannot be migrated.
Example 3
Finding:
The average dwell time on S1 or S2 is short.Reason:
If, according to the migration procedure, all files which have been unused for X days are migrated, the number X has been set too low.Remedy:
If a higher number is specified for X, the storage space used on S1 is increased, as is the length of time before a recall. The number of recalled files is reduced and the migration becomes more efficient overall.
Example 4
Finding:
The average time between generation of a recall request and the start of its processing is too long.Reason:
The number of HSMS subtasks is probably too small.
Example 5
Finding:
The average time during which the magnetic tape cartridges are mounted is too long.Reason:
Too many files are being recalled.