Most of the conditions for good performance, established for the
time-shared batch systems, apply equally well to mixed mode systems.
However, two major factors make any determination more difficult to
make. First, get evidence to show that, in all circumstances, priority
is given to maintaining good interactive response, and that nontrivial tasks truly take place in the background. Second, background tasks, no
matter how large, inefficient, or demanding should not be allowed to
dominate the overall utiliZation of the time-sharing system. In other
words, in mixed mode operation, get evidence that users with poor
characteristics are discriminated against for the sake of maintaining a
healthy system for the remaining users.
A number of other conditions are more obvious and straightforward. You need to measure response and determine at what point it becomes
unacceptable and why. Studies of time-sharing systems have shown that a
user's rate of working is closely correlated with the system response. When the system responds quickly, the user is alert, ready for the next
interaction, and thought processes are uninterrupted. When the system
response is poor, the user becomes sluggish.
For interactive environments, a need exists to analyze command usage.
Average execution time of the truly interactive commands can provide
data for validation of the Queue 1 execution time.
126 IBM 1M/370 System Programmerls Guide

April 1, 1981
Accounting Records
The accounting data gathered by VM/370 can help in analysis of overall
system operation. Also, accountinq data can be used to bill users for time and other system resources they use.
There are three types of accounting records: the virtual machine user
records, records for dedicated devices as well as T-disk space assigned
to virtual machine users, and accounting records generated as a result
of user initiated DIAGNOSE X'4C' instruction. A eMS batch virtual
machine creates an accounting record with the userid and account number
of the user who sent his job to the batch machinee Accounting records
are prepared as SO-character card images and sent to a punch file at
various times. output class C
is reserved for accounting records.
If the amount of free storage (available page frames) is relatively
small and the card punch is not periodically assigned to punch CP's
accounting cards, it is possible for CP's accounting routine to
progressively use a significant percentage of the available page frames
and cause a paqe thrashing condition to occur in VM/370. This happens
because the accounting routine creates and maintains accounting records
in real storaqe, and does not free that storage space until the
accounting records are punched on the real system card punch.
To eliminate this problem, it is recommended that one punch pocket be
permanently dedicated to this accounting function, or if that is not
feasible, to punch all the accumulated records every 1 to 2 hours.
Accounting cards are punched and selected to pocket 2 of any class C
card punch when a user logs off of the system, detaches a dedicated
device or T-disk, or issues a DIAGNOSE code X' 4C' instruction. (If the
real punch is a 2540, the accounting cards are put in pocket 3.) These
records should be kept for system accounting purposes.
Accounting Records for Virtual Machine Resource
Usage
The punched in the accountinq card when a user ends his
terminal session (or when the ACNT command is invoked) is as follows
(columns 1-28 contain character data; all other data is in hexadecimal
form, except as noted): 1- 8
9-16
17-28
29-32 33- 36 37-40 41- 44
45- 49-52
53-56
57- 60 Contents Userid-- Account num1::er Date and Time of Accou nting (mmddyyhhmmss)
Number of seconds connected to VM/370 System of processor time used, including time
for VM/370 supervisor functions
Milliseconds of virtual processor time used
Number of page reads
Number of page writes
Number of virtual machine SIO instructions for
nonspooled I/O Number of spool cards to virtual punch
Number of spool lines to virtual printer (this
includes one line for each carriage control command)
Part 2. Control Program (CP) 127