For detecting resource overloads the following data is required:
workload values: CPU, channels, devices (disks), main memory
With disks, SM2 sees exclusively logical volumes.number of I/O operations per device
number of tasks per device queue
Working set requirement for each task category
To make clear the dependence of response time behavior on resource workload, it is helpful to maintain the following statistics in parallel:
Connection-specific response time statistics (monitoring program BCAM-CONNECTION)
The use of the following monitoring programs is also recommended:
Extended system statistics (monitoring program SYSTEM)
Among other things, this monitoring program provides information on dilations and dwell times in the system queues for each category, as well as information on the hardware service time (hardware duration) and software service time (software duration) for DMS and paging I/O operations.Extended device statistics (monitoring program SAMPLING-DEVICE) The monitoring program is started automatically when SM2 is called, but the hardware and software service times for each volume are not recorded. If monitoring is also to include these values, the monitoring program must be restarted with the following statements:
//SET-SAMPLING-DEVICE-PARAMETER DISK-SERVICETIME=*ON//CHANGE-MEASUREMENT-PROGRAM TYPE=*SAMPLING-DEVICE.Channel statistics (monitoring program CHANNEL-IO)
This monitoring program is used to ascertain the number of I/O operations and the number of PAM blocks for each channel.Catalog statistics (monitoring program CMS)
This monitoring program provides information on the number, type and duration of the accesses to catalog entries on each pubset.VM2000 statistics (monitoring program VM)
In VM2000 operation this monitoring program should only be started in the monitoring system (but there by default). In the monitoring system it supplies the CPU share of all guest systems and on /390 servers the share of hypervisor active and idle.