Previous Topic: 2.2.2 ISC and MRO Environment

Next Topic: 2.2.4 CICS CPU Time

2.2.3 CICS Response Time


The ideal measure of response time is the end-to-end
transaction response time, which starts when the terminal
user enters a request and ends when the result of the request
is displayed at the terminal.  Unfortunately, this type of
response measure requires tracking all events occurring at
many different points of the communication network.
Therefore, it is not available from any of the CICS monitors.
Response time in terms of CICS is the internal transaction
elapsed time.


MEASURING CICS RESPONSE TIME

The CICS internal transaction response time is the time
between the CICS initiation of a task and the CICS
termination of the task.  This measure includes two primary
components:   dispatch time and wait time.  Dispatch time
includes CPU time and the time CICS is involuntarily
interrupted while the task is dispatched.  Wait time includes
the time CICS is in a wait state, such as dispatch queue
wait, I/O waits, and VSAM file string waits.  Certain wait
states cause the CICS task to be suspended, the most notable
of which is the wait for terminal input.

The CA MICS Analyzer Option for CICS normally derives
response time based on the transaction start and end
timestamps as provided by the CICS monitors.  However, for
conversational tasks, it further excludes user think time
from the transaction response time to prevent skewing of the
internal response measure.  The actual response derivation is
described below.

MEASURING CONVERSATIONAL TRANSACTION RESPONSES

A conversational task, as described earlier in this chapter,
involves multiple interactions between the terminal user and
CICS.  To provide a more accurate response time for this type
of transaction, some CICS monitors subtract the time the
transaction is suspended in CICS while waiting for terminal
input from the transaction elapsed time.  ASG-TMON for CICS
TS (TCE) excludes terminal wait time for conversational tasks
by generating a transaction record for each segment (a pair
of terminal I/Os) of the conversation.  The same is true for
CMF when the MNCONV=YES parameter is specified in the SIT (at
CICS/ESA 4.1 and higher), or the CONV=YES parameter is
specified in the DFHMCT TYPE=RECORD macro (pre-CICS/ESA 4.1).
Otherwise, CMF writes one transaction record for the entire
conversational task, in which case the elapsed time includes
user think time.

To ensure the CA MICS response derivation accurately reflects
the internal response time, the CA MICS Analyzer Option for
CICS subtracts the minimum of either terminal control wait
time or suspend time from transaction elapsed time.

internal response=end-start-MIN(suspend,TC wait)

Terminal control wait time includes the time waiting for
terminal input, but it also includes time waiting for other
terminal control services that might not cause the task to be
suspended.  Suspend time includes time the task waited for
terminal input and any other suspend-eligible reasons.  One
frequent cause of task suspension is having CICS operate at
the active max task limit.  For this reason, suspension has
nothing to do with terminal control.

Figure 2-3 shows the flow of control for one CICS
transaction, including waits for transaction-required
resources during processing.  An example of such a
requirement is a CICS Terminal Control read operation.

Comment Terminal CICS Terminal CICS Task User User Control Control Program ---------- ---------- ---------- ---------- ---------- hit ENTER | wait for | network +------------+ | recognize request | +-------------+ _ | | initiate | task | | | create internal task response | time +-----------+ waits for | | resources | process | | | +-----------+ | | | terminate | task - | +-------------+ final terminal wait for output network +--------------+ | output shown | _

  Figure 2-3.  CICS Control Flow for a Transaction