Browse Prior Art Database

Method and apparatus for z/OS PTF number determination

IP.com Disclosure Number: IPCOM000198697D
Publication Date: 2010-Aug-12
Document File: 2 page(s) / 28K

Publishing Venue

The IP.com Prior Art Database

Abstract

In order to package fix packs on z/OS systems there are certain size restrictions in place that prevent from building in a single PTF package. There is a size limit on the number of records as well as one on the number of parts. In order to determine the number of packages that is to be used to complete a fix pack build (PTF build) old tools implemented a simple greedy algorithm that produced inadequate results. It also required all parts be transmitted from MVS to VM as part of the process. The transmission added unnecessary time to the number determination. This primitive solution took first the record size restriction then the part size restriction and gathered parts until those were met. This often resulted in a very obvious need to merge two PTFs that were not automatically allocated more optimally by the automation tooling. Merging PTFs often resulted in errors that led to re-builds and build delays. Therefore, a more efficient way of determining the proper distribution of parts is needed to avoid the risk involved in merging PTFs and facilitate and expedite the building process.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 49% of the total text.

Page 1 of 2

Method and apparatus for z/OS PTF number determination

This new algorithm is using a total different approach by doing the following:
1. Don't need to send parts from MVS to VM anymore.
2. Check every part's data type and size and also check F80 data sets if they exist and use F80 data sets instead to have more accurate parts info.
3. Check parts total for fitting the limits for each PTF can contain.
4. Then finally calculate the total numbers of PTF needed.

The performance compared with old one is about the same but the new one produces more accurate and reliable results in email.

Example of result.

NEW TOOLING:
Splitting parts into ptf sets of less than 1580000 FB80 records ...

Wrote AUX1 with 880 parts, 1568283 records

Wrote AUX2 with 880 parts, 1552646 records
Wrote AUX3 with 880 parts, 1574474 records
Wrote AUX4 with 880 parts, 1577470 records
Wrote AUX5 with 650 parts, 1574758 records
Wrote AUX6 with 95 parts, 790411 records
Wrote AUX7 with 1 parts, 2167614 records
Wrote AUX8 with 1 parts, 2050686 records
Wrote AUX9 with 1 parts, 1743075 records
AUXSPLIT processing complete.

Made 9 apars on this run.

There were 3 large-part apars.

READY

OLD TOOLING:


It was found 4268 parts that were divided as follows:
TOTALS FOR AUX1 (PTF #1): # parts= 6 # FB80 records= 1580000 TOTALS FOR AUX2 (PTF #2): # parts= 14 # FB80 records= 1580000 TOTALS FOR AUX3 (PTF #3): # parts= 33 # FB80 records= 1580000 TOTALS FOR AUX4 (PTF #4): # parts= 110 # FB80 records= 1580000 TOTALS FOR AUX5 (PTF #5): # parts= 497 # FB80 records= 1580000 TOTALS FOR AUX6 (PTF #6): # parts= 880 # FB80 records= 448930 TOTALS FOR AUX7 (PTF #7): # parts= 880 # FB80 records= 168688 TOTALS FOR AUX8 (PTF #8): # parts= 880 # FB80 records= 102899 TOTALS FOR AUX9 (PTF #9): # parts= 880 # FB80 records= 21807 TOTALS FOR AUX10 (PTF #10): # parts= 85 # FB80 records= 243 BBO03226 02167149
BBO03224 02050245
BBO12488 01742699

The first part of the improved method consists in performing computing operations without the need to transmit parts from one side of the system to another. The main reason this had to be done in the previous method was because on VM the datasets were transformed from variable block 255 format onto fix block 80. The conversion presented a challenge since if a record's actual size for instance on the variable block format was 90 it would take 2 records once converted. However, if it's actual size was 250 it would take 4 size 80 records when converted. The conversion also added additional space which made it very difficult to estimate the accurate final number of records unless processing took place after the

1

Page 2 of 2

conversion. The solution to this was to utilize converted datasets on the MVS side that were alread

block 80 to record each part's number of FB80 (fix block 80) records. The first part of the tooling therefore consisted in generating a list of all parts involved in the PTF build and their FB80 record size that would later be used by the second phase of the new method...