Chapter 26: Further Reading

Batch Performance at Scale: I/O Optimization, Buffer Tuning, SORT Optimization, and DFSORT Tricks

IBM Documentation

  • DFSORT Application Programming Guide (SC23-6878) The definitive reference for everything in Section 26.3. Chapters on INCLUDE/OMIT, OUTREC/INREC, OUTFIL, and ICETOOL are essential reading. The "Examples" appendix contains over 200 worked examples covering sort, merge, copy, reformat, split, join, and report operations. If you read one reference from this list, make it this one.

  • DFSORT Installation and Customization Guide (SC23-6881) Covers DFSORT runtime options including MAINSIZE, HIPRMAX, DYNALLOC, and FILSZ. The section on installation defaults explains why your shop's DFSORT may not be using optimal settings — and how to change them. Also covers DFSORT's zIIP offload configuration.

  • z/OS MVS System Management Facilities (SMF) Reference (SA38-0667) Complete field-level documentation for all SMF record types. For this chapter: Type 14/15 (dataset activity), Type 30 (job accounting), Type 42 (DFSORT statistics), Type 101/102 (DB2 accounting). The field layouts and interpretation notes are essential for building the performance dashboards described in Section 26.6.

  • z/OS RMF User's Guide (SC34-2664) RMF report interpretation for CPU activity, channel activity, device activity, paging activity, and workload activity. The "Performance Analysis" section provides guidance on identifying bottlenecks — channel saturation, device contention, and WLM throttling — that SMF alone can't reveal.

  • Enterprise COBOL for z/OS Programming Guide (SC27-1411) Chapter on compiler options covers OPT levels, FASTSRT, NOSEQCHK, NUMPROC, TRUNC, and SSRANGE in detail. The "Performance Tuning" chapter specifically addresses COBOL-level optimization techniques including generated code analysis. The LIST output interpretation guide is invaluable for Section 26.4's generated code discussion.

  • Enterprise COBOL for z/OS: Performance Tuning Guide (GC27-9202) Dedicated IBM publication covering COBOL batch performance. Includes benchmark comparisons for OPT levels, FASTSRT impact measurements, and guidelines for data definition choices (COMP vs. COMP-3 vs. DISPLAY) that affect CPU efficiency.

  • z/OS DFSMS: Using Data Sets (SC23-6855) Comprehensive coverage of access methods (QSAM, BSAM, VSAM), buffering, and BLKSIZE optimization. Chapter on sequential access performance includes the track capacity tables and half-track blocking formulas referenced in Section 26.2.

  • z/OS DFSMS: Access Method Services for Catalogs (SC23-6853) VSAM cluster definition, including BUFFERSPACE, CONTROLINTERVALSIZE, and FREESPACE parameters that affect batch VSAM performance. The LISTCAT output interpretation helps verify VSAM buffer and CI size settings.

  • DB2 12 for z/OS: Performance Monitoring and Tuning Guide (SC27-8850) Chapters on batch performance, commit frequency analysis, sequential prefetch, list prefetch, and partition-level parallelism. The IFCID trace record documentation enables the DB2 performance analysis described in Section 26.5.

  • z/OS MVS Programming: Authorized Assembler Services Guide (SA23-1371) For understanding Hiperbatch (Data Lookaside Facility) and data-in-memory at the system programming level. Chapter on DLF configuration covers the COFDLFxx parmlib member syntax and object registration.

IBM Redbooks

  • Bob Rogers, "z/OS Batch Processing" (SG24-6698) End-to-end batch processing on z/OS — initiator management, JES2/JES3 processing, SMF recording, and batch performance measurement. The performance analysis methodology maps directly to the approach described in Section 26.1.

  • Paolo Bruni et al., "DB2 11 for z/OS: Performance Topics" (SG24-8222) The batch-specific chapters cover commit frequency analysis, lock escalation monitoring, sequential and list prefetch tuning, and partition-level parallel query execution. Case studies include before/after measurements similar to CNB's performance project.

  • Octavian Lascu et al., "IBM zEnterprise System Technical Guide" (SG24-7962) Hardware architecture relevant to batch performance — CP engine types (GP, zIIP, zAAP), cache hierarchy, channel architecture, and FICON connectivity. Understanding the hardware helps explain why certain optimizations (half-track blocking, sequential prefetch) produce the improvements they do.

  • Frank Kyne et al., "ABCs of z/OS System Programming: Volume 10" (SG24-6987) Covers SMF configuration, RMF setup, and performance monitoring infrastructure. Essential background for Section 26.6's discussion of SMF and RMF-based performance analysis.

Books

  • David Shelby Kirk, "MVS Performance Management" (McGraw-Hill) Classic text on z/OS performance engineering. Chapters on I/O subsystem performance, access method tuning, and WLM interaction with batch workloads remain relevant despite the book's age. The analytical framework for decomposing elapsed time into CPU, I/O, and wait components is the foundation for Section 26.1.

  • Thomas Kern and David Shelby Kirk, "z/OS Performance: Monitoring and Tuning" (IBM Press) Modern successor to Kirk's earlier work. Covers hardware performance counters, RMF III monitoring, and WLM-managed batch workloads. The chapter on zIIP and zAAP exploitation is particularly relevant to Section 26.7.

  • Jim Tate et al., "Introduction to Storage Area Networks and System Networking" (IBM Redbooks, SG24-5470) Understanding the storage subsystem — FICON channels, DS8000/DS8900F controllers, caching algorithms — explains why BLKSIZE and sequential access patterns have such dramatic performance impact. The cache architecture chapter reveals why sequential workloads achieve 95%+ cache hit ratios.

Online Resources

  • IBM DFSORT Tricks (Frank Yaeger's Site) Frank Yaeger, the principal architect of DFSORT for decades, published extensive examples of DFSORT/ICETOOL techniques that go far beyond sorting. DFSORT can perform joins, lookups, generation of test data, complex reformatting, and conditional processing. Many of the ICETOOL patterns in Section 26.3 derive from techniques Frank documented.

  • SHARE Conference Proceedings — Batch Performance Track SHARE conferences regularly feature sessions on batch performance optimization from practitioners at major enterprises. Search the SHARE proceedings archive for "batch performance," "DFSORT," and "I/O optimization" for real-world case studies similar to CNB's and FBA's.

  • IBM z/OS Hot Topics Newsletter Published by IBM's z/OS development team, Hot Topics often covers performance tuning techniques, new DFSORT features, and zIIP exploitation strategies. The articles are written for practitioners, not marketing — they include real JCL and control statement examples.

Chapter Cross-References

  • Chapter 4 (Dataset Management) — BLKSIZE fundamentals, SMS data class configuration, VSAM cluster design. The I/O optimization in this chapter builds directly on Chapter 4's dataset architecture decisions.

  • Chapter 6 (DB2 Optimizer) — Access path selection, EXPLAIN analysis, catalog statistics. The DB2 batch performance patterns in Section 26.5 extend Chapter 6's optimizer concepts to batch-specific scenarios.

  • Chapter 23 (Batch Window Engineering) — Critical path analysis, job dependency graph, throughput math. Chapter 23 defines the window; Chapter 26 optimizes the jobs within it.

  • Chapter 25 (Parallel Batch Processing) — Parallel job design, partition-by-key strategies. Chapter 25's parallelization creates the architecture; Chapter 26's tuning maximizes each parallel stream's throughput.

  • Chapter 27 (Batch Monitoring) — Operational monitoring, alerting, trend analysis. Chapter 27 builds the ongoing monitoring framework that sustains the performance improvements achieved in Chapter 26.

  • Chapter 29 (Capacity Planning) — MSU forecasting, storage growth, transaction volume modeling. The performance baseline established in Chapter 26 feeds directly into Chapter 29's capacity projection models.