IBM released the newest version of its flagship operating system z/OS 1.13 in September 2011. For developers pondering a migration to this version, there are three major enhancements worth noting.
Unix file system relief for zFS
The Unix System Services (USS) file system zFS and its predecessor, HFS, has been slowed by performance and operational issues, but z/OS 1.13 attempts to speed things along.
Before z/OS 1.13, if zFS found a corrupted file system it would ask the operating system to restart. When it reinitialized not all the file systems were remounted, and there was no guarantee others weren’t corrupted. The update puts zFS through a more extensive process in which it will stop incoming requests, restart its kernel subtask and internally remount all file systems. If all goes well the files will be available and uncorrupted. If things don’t go well users will have to revert to restoring and remounting the damaged file system along with a possible Initial Program Load (IPL).
The z/OS 1.13 update attempts to solve problems with zFS's I/O performance. In previous versions, each file system had a designated owner that performed all update activities for files within the mount. Other systems wanting to update something in the file system had to talk to the owner through cross-coupling facility (XCF) which appreciably increased the I/O time. The z/OS 1.13 update didn’t remove file system owners, but now any logical partition (LPAR) can read and update a zFS file directly. IBM says customers can expect a 50% to 150% improvement in I/O performance. Updates to file metadata are still done through the owner.
RELATED z/OS LINKS
The downside, IBM warns, is that zFS will use more disk space since z/OS 1.13 stores data in 8K blocks.
z/OS 1.13 boosts Java performance
Although IBM has been touting Java on the mainframe for a long time, adoption has been slow in part because of poor integration between Java and traditional mainframe languages like COBOL. In one particularly glaring example, COBOL and Java programs running in the same address space couldn’t coordinate database updates because each used different interfaces to DB2. But z/OS 1.13 introduces z/OS Batch Runtime, meant to both boost Java performance and make it cozier with COBOL.
To get things rolling, IBM offers a new startup program, BCDBATCH, a Java application that can call a COBOL or Java main program. BCDBATCH uses JZOS to initialize Java virtual machine (JVM) and launch the Java environment.
Batch Runtime also allows COBOL and Java to coordinate updates to DB2, so changes are committed or backed out together. This coordination requires registration of the job with z/OS' Resource Recovery Services (RRS).
Batch Runtime currently only works with the 31-bit JVM. It would be nice if support extended to other mainframe languages such as Programming Language One (PL/1) as well as coordinated units of recovery with Information Management System (IMS).
Administration tool z/OSMF gets under-the-hood tweaks
IBM continues to enhance z/OSMF to make it the administration tool of choice for mainframe systems for a few reasons. First, z/OSMF provides a familiar interface for fresh faces who may not immediately grasp Interactive System Productivity Facility (ISPF). Second, it has the potential to greatly simplify the administration and deployment of z/OS systems, though possibly at the risk of lowering the skill level of said administrators.
A new z/OSMF feature is capacity provisioning. At this point z/OSMF is limited to monitoring the status of facilities like Capacity on Demand and interfacing with Resource Management Facility (RMF) and Workload Manager (WLM) for performance information. IBM says this integration makes capacity decisions quicker and more accurate, although executing any workload related changes still requires the Windows-based program Capacity Provisioning Control Center (CPCC).
z/OSMF software deployment is supposed to provide rigor around System Modification Program/Extended (SMP/E) managed software, both for IBM and independent software vendors (ISVs). It can also clone software within the same systems complex (sysplex) or across sysplexes and manage software maintenance through full replacement instead of updating in place.
To facilitate deployment, z/OSMF has a checklist for the administrator. As the systems programmer ticks items off the list, z/OSMF generates and submits jobs to perform the actions.
You'd think software deployment is most useful for installations with low experience levels or a lot of products to maintain with too few people, but automated software management probably works best in plain, vanilla shops. Highly customized installations, either through system changes or unique naming conventions, may require more time fighting the tool than using its features.
More changes for z/OS
At the spring 2012 SHARE conference, IBM said it plans to move z/OS to a biannual release schedule, because most customers couldn’t keep pace with the annual schedule with only a small percentage able to stay current. However, between releases, IBM intends to publish new functionality through Small Product Enhancements (SPEs).
In other news, an IBM direction statement indicated the company would issue version 2 of z/OS in the second half of 2013.
About the expert: Robert Crawford has been a systems programmer for 29 years. While specializing in CICS technical support he has also worked with VSAM, DB2, IMS and assorted other mainframe products. He has programmed in Assembler, Rexx, C, C++, PL/1 and COBOL. The latest phase in his career finds him an operations architect responsible for establishing mainframe strategy and direction for a large Insurance company. He lives and works with his family in south Texas.