Queue time delays: EQT + Delay Time = AQT


Effective Queue Time (EQT) is the time a job waits before an initiator considers it for selection. It is a measure of how long jobs are waiting for CPU resources - CPU resources might vary depending on the constraints in effect at the time. Actual Queue Time (AQT) is how long a job is in the queue before it is initiated.

Delays account for the difference between the EQT and AQT. More than one reason might account for a delay. ThruPut Manager AE attributes delay time to a single reason as long as that reason persists. Then any remaining delay time is attributed to the next reason. Here's a subset of the common reasons in "attributable order":

  • JTS, User, DJC - These delays are user-initiated and are all techniques to delay a job starting until a trigger occurs to release them. Though these jobs have been submitted they do not age until the trigger occurs.
  • Datacenter - These delays are datacenter-initiated and include systems offline, regular affinity delays, JAL errors and so on. Jobs age without restriction.
  • Operations - These delays are Operations-initiated, such as when they place a HOLD on the job for some reason. The job continues to age through the hold.
  • DFSMShsm Recalls- These delays occur when a job is waiting for recalls to complete. The job continues to age through the recall.
  • Virtual Volume Staging - These delays occur when a job is waiting for recalls to complete. The job continues to age through the delay.
  • Scheduling Environment - These delays occur when the job is waiting for the appropriate Scheduling Environment to become available. The job continues to age through the delay.
  • Binding - These delays occur when the job is waiting for the appropriate combination of binding agents to become available. The job continues to age through the delay.
  • DCS - These delays occur when a job is waiting for a data set that is held by another job. GS jobs do not age during this time. PS jobs do.


Queue_Time_Graph.JPG

The area below the EQT line reflects how long jobs are waiting for CPU resources. The system was busiest at about 5:30. The area between the lines reflects the delays. Where the lines are close together, there is little delay; the greatest delays occurred at about 2:30.

Whatever the cause of the delay, if the job continues to age through the delays, then Serviced Delivered field reflects the where the AQT was with respect to the target, acceptable and critical thresholds. If the delay causes the job to stop aging, then the Service Delivered column reflects where the job would have been selected without that delay. For instance, if a GS job is held for data set contention (with a Target service time of 12 minutes) with a hold and releases it after 21 minutes and 2 seconds, and the job is initiated 31 seconds later, with no further delays, the EQT is 31 seconds, the AQT is 21 minutes and 33 seconds, the DCS delay is 21 minutes and 2 seconds, and the Service Delivered is at target.

Exceptional jobs have a value of Indeterminate in the Service Delivered field. For more information see Exceptional Jobs.

******************************* SLM Job Summary ****************************
* Read on Node PARIS, System DOCT Executed on Node PARIS, System DOCT *
* Read at 10:54:33 Recognized by SLM at 10:54:33 *
* Elapsed time from when job was read in to SLM recognition... 00:00:01 *
* Service Policy Information: *
* Policy Name: MIXED07 Description: Work with PS & GS Groups*
* Activated at 10:12:32 Last Modified at 12:59:34 *
* Service Group Information: *
* Control Center: BNK Description: Adhoc banking jobs *
* Type : QA Description: Pre-prod testing *
* Service Mode: STANDARD *
* Queue Time Information: *
* Actual... 00:01:33 Delay... 00:01:20 Effective... 00:00:13 *
* Target... 00:05:00 Service Delivered: Before Target *
* Queue Time Delays *
* Datacenter.............. 00:00:00 Operations............ 00:00:00 *
* JTS..................... 00:00:00 JSS....................00:00:00 *
* Scheduling Environment.. 00:00:00 User...................00:00:00 *
* Binding/Environment..... 00:01:20 Limiting.............. 00:00:00 *
* DJC..................... 00:00:00 DBS................... 00:00:00 *
* Resource Conflicts...... 00:00:00 DCS................... 00:00:00 *
* Recalls................. 00:00:00 Virtual Volume Staging.00:00:00 *
* Times Restarted... 0 *
* System Status at time of Job Selection: *
* Service Class: SLMSTDMP Color: Green *
* Performance Index: 0.7          Delay: 71%    CPU available: 21% *
* CPC busy: 41.97% MVS busy: 3.73% LPAR busy: 2.48% System color: Green *
*
* Execution Time Information: *
* Job selected at 10:56:07 *
* Elapsed... 00:02:28 Delay... 00:00:01 Effective... 00:02:27 *
* CPU Service Units... 456,863 *
* Effective CPU SU Rate... 3,107.91 SUs/Sec *
* Execution Time Delays *
* Initiation... 00:00:00 DBS... 00:00:00 DCS... 00:00:00 *
* Recalls... 00:00:00 Allocation... 00:00:01 *
* Summary *
* Elapsed Times *
* Pre-SLM................ 00:00:01 *
* Queue *
* Delay................ + 00:01:20 *
* Effective............ + 00:00:13 *
* Lost due to Restarts... + 00:00:00 *
* Execution *
* Delay................ + 00:00:01 *
* Effective............ + 00:02:27 *
* Actual Elapsed Time.... = 00:04:02 *
* *
***************************************************************************

In the Job Summary Report Sample above, this job received service as a BNK QA Service Group; the service delivered is "before target"; there were two delays - 1 minute, 20 seconds for binding delay at queue time, and .01 seconds allocation at execution time.

Note

Sections of the Job Summary report are also presented in response to a "SLM DISPLAY JOB" command and TM/UDF.


 

Tip: For faster searching, add an asterisk to the end of your partial query. Example: cert*