S14001 Visa Data Coll User Message Count Support Merged

TPFUG_Req_Number: S14001
ORIGIN DATE: 12/16/2013

REQUIREMENT ABSTRACT

User customization of data collection/reduction should be enhanced to support
user message counters and the ability for the user to define how to calculate weighted message rate.

REQUIREMENT DESCRIPTION

Currently 5 messages counters (high speed processed, low speed, high speed routed, created, and SSCP) and the NSD count of weighted IP messages are written at the start/end of data collection and by the system data control.

The calculation for weighted messages is :
(high speed processed – low speed ) + (low speed * Websphere Message Broker(WMB)) +
(high speed routed * WT2) + weighted IP messages

The user can only control the weighting of low speed and high speed routed via the Websphere Message Broker(WMB) and WT2 controls in TPF/c_dcdr.h that are also maintained in file system file etc/dataco.txt.

The IBM message counters do not necessarily describe all the message count types of interest with sufficient granularity and the weighted message calculation may not describe with sufficient granularity how the customer wants to measure weighted message rate. This may be particularly true when a single port supports messages with very different resource consumption profiles. In order to tailor message counts and the weighted message calculation, the user currently has to apply local mods to repurpose or massage the content of the IBM message counters and/or change the offline data reduction code to calculate weighted messages as required.

To avoid the need for these local mods, IBM should:
a) Provide an array of at least 16 counters that the user can use to count messages of various types according to their needs. These user counters should be written out with the IBM defined counts and should appear in the SYSTEM SUMMARY section of the data reduction report so that min/max/mean values are shown. Support for user-defined descriptions to be used in data reduction for the in-use user counters is required.

B) Provide the means for the user to define the weighted message rate calculation to be used by data reduction. The weighted message calculation should allow for inclusion of IBM counters and user-defined counters.


CRITICAL SUCCESS FACTORS

User defined message types must be collected by data collection and reported on in data reduction, and the weighted message calculation must be definable by the user without having to change any code in data collection or reduction.

Changes to user definitions should be supported without the need for an IPL.


ALTERNATIVE SOLUTIONS



SOLUTION CONSIDERATIONS
-
Currently, user modification of data collection/reduction is achieved through use of definitions in TPF/c_dcdr.h and the Posix file in etc/dataco.txt. ZMEAS SET is used to manage user-defined settings.

This infrastructure could be used to support the new requirements. Description definitions to be used by data reduction for each in-use user message count field could be added along with the weights to be applied to IBM and in-use user defined message counters. Assigning weights per IBM/user counter field would allow the weighted message calculation to be defined by the user in a similar manner to today.

For example, the weighted message calculation could be defined as (Wx is the weight for the specific counter type in dataco.txt):
((high speed processed *W1)- low speed)) + (low speed * W2) +
(high speed routed * W3) + (weighted IP messages * W4) +
a(user counter1 * W5) + (user counter2 * W6)…. + (user counterN * Wn+4)

A weight of 0 must be allowed in order to exclude a message count type. This approach would support customers who needed to change the hardcoded vanilla calculation used today, and those who do not.

ZSTAT cvoff0.asm source need not be updated to display the user counts. If the user wants to display them they can do so via the provided user exit usta.asm. Similarly, CDC may not need to accommodate these counters since user exits are already supplied.