Blogs

BITMARCK grows service to help reduce utilities costs

DB2 for z/OS Systems Programmer, BITMARCK Technik Ltd.

Many corporations today face the challenge of growing the business while maintaining a competitive edge and controlling or reducing IT spending. Enterprises deploying IBM DB2 for z/OS databases are experiencing significant growth in data volumes and complexity. For DB2 for z/OS applications with high-transaction activities involving insert, delete and update operations, there can be rising costs associated with DB2 utility processing. This situation is especially true for LOAD, REBUILD INDEX and REORG DB2 utilities. The sort processing within these utilities tasks can lead to increased central processor unit (CPU) usage and elapsed time. 

Traditionally, DB2 Utilities called the system sort and used the standard application programming interface (API) that every other sort-invoking program uses. Anyone familiar with the inner workings of DB2 Utilities knows that it is not like every other sort invoker. What is called for is tight collaboration with utilities and sort to provide a high degree of performance and availability. This collaboration also attempts to reduce costs and take advantage of specific IBM z Systems platform features such as the System z Integrated Information Processor (zIIP).

Enhancing the use of system resources

Developers from the DB2 Utilities group at the IBM Silicon Valley Lab partnered with Syncsort, who specializes in big-iron-to-big-data solutions that collect, integrate, sort and distribute data rapidly, with reduced resources and cost, to jointly develop DB2 Sort for z/OS. They worked to build a sort solution that is specifically designed to operate in the distinctive environment created by DB2 Utilities. DB2 Sort for z/OS takes advantage of algorithms and optimizations that were developed for the way DB2 Utilities operates. It incorporates parallel processing and zIIP offload in ways that are different than using a generic API (see figure). As a result, DB2 Sort can reduce utility processor time by up to 60 percent and elapsed time by 45 percent, depending on the specific hardware and software environment and usage. 

The primary goal of DB2 Sort is to help reduce costs through improved performance. Depending on each organization’s specific needs, DB2 Sort may help in different ways. For some, the enhanced processor time may allow for a delay in the next hardware upgrade. Other customers may find value in the reduction of a utility maintenance window, which can improve application availability and help an organization achieve successful outcomes.

Another goal of DB2 Sort is efficient use of resources. Sorting can be a resource-intensive process, but there is a finite limit on those resources. DB2 Sort takes a global view and divides the available resources in a way that benefits the overall utility performance instead of looking at each process individually. For example, at times it makes sense to reduce one process of resources to give a higher-priority process added resources. This level of control is critical to helping achieve a highly successful overall performance.

DB2 Sort is also designed to intelligently manage all resources that it uses while requiring minimal input from line-of-business users. This approach is evident in the way it manages the amount of memory it uses both for virtual storage and memory object space, as well as the way it dynamically allocates and manages the SORTWK DASD space. End users need not be concerned with setting a series of parameters that limit what resources can be used—and in turn limiting the performance that could be achieved—or worry about job failures because of under-allocated resources. DB2 Sort is designed to be easy to manage and use.

Seizing an opportunity to grow a customer base

BITMARCK Technik GmbH, a subsidiary of BITMARCK Holdings GmbH, is a German provider of IT services for outsourcing and portal solutions in the healthcare insurance industry. This enterprise serves as an example of an organization that achieved success in controlling or reducing IT spending by experiencing heightened performance efficiency for DB2 Utilities sorting. 

BITMARCK has over six million customers and is a strong IT partner of the Social Health Insurance (SHI), a key branch of the German social security system. BITMARCK’s challenge was finding ways to grow the number of customers it services with little if any additional costs to its IT infrastructure and staff. 

The organization has over 60 DB2 for z/OS data management environments in 10 logical partitions (LPARs) with many applications, including SAP. BITMARCK wanted to reduce costs in its large and growing environment, especially for long-running LOAD, REBUILD INDEX and REORG jobs.

Efficiency challenge: In its desire to help reduce the costs of running the DB2 Utilities suite, BITMARCK realized that sorting was a large component of jobs using DB2 Utilities; therefore, improving the sorting process was a good place to start. The organization previously participated in an IBM beta program for DB2 Sort for z/OS Version 1.3. By procuring and installing DB2 Sort, enhanced processor and elapsed time savings were achieved along with overall enhanced use of the DB2 for z/OS maintenance window. 

With DB2 Sort deployed, BITMARCK experienced immediate relief during sorting jobs by DB2 Utilities. The organization saw an overall immediate 40 percent reduction in processor usage. In addition, it experienced enhanced zIIP offload, and helped reduce costs by delaying hardware upgrades. Impressed with the ease of use provided by DB2 Sort and these results, BITMARCK agreed to participate in the IBM beta program for DB2 Sort Version 2.1.

The testing program: The IBM and Syncsort partnership that resulted in DB2 Sort for z/OS developed a program that enables organizations and potential customers to test DB2 Sort in their environments. The installation verification program (IVP) offers a self-contained evaluation methodology to demonstrate the value of DB2 Sort. The IVP verifies that DB2 Sort is installed and running properly, and it performs testing with the following characteristics in the organization’s environment: 

  • Practical: Efficient completion of tests in a reasonable time frame, with no additional challenges for IT staff
  • Representative: Accurate creation of models that approximate relevant production data
  • Reproducible: Establishment of reliable and repeatable evaluation results

Some of the results of BITMARCK’s IVP from the DB2 Sort 2.1 beta testing show the LOAD, REBUILD INDEX and REORG jobs with various data volumes and number of indices (see table). These results reflect improvement in all areas for sort and utility processors, elapsed time and zIIP offload. For BITMARCK, these results mean continued savings using DB2 Sort.

Object size, in billions of bytes

DB2 utility

Sort processor time

Step processor time

Step elapsed time

DB2 Sort zIIP offload

8

LOAD

78.1%

37.3%

22.3%

60.1%

REBUILD INDEX

76.1%

47.7%

39.4%

60.0%

REORG

45.0%

41.6%

30.5%

25.4%

10

LOAD

77.5%

37.0%

21.3%

59.5%

REBUILD INDEX

75.4%

46.6%

39.6%

60.1%

REORG

45.3%

41.7%

31.0%

25.0%

25

LOAD

79.1%

48.8%

37.0%

58.4%

REBUILD INDEX

75.3%

47.9%

43.3%

57.3%

REORG

54.4%

50.2%

38.6%

27.4%

40

LOAD

73.8%

42.3%

39.6%

56.8%

REBUILD INDEX

75.9%

48.9%

44.7%

56.7%

REORG

55.0%

50.8%

40.7%

26.6%

Averages

67.6%

45.1%

35.7%

47.8%

 

BITMARCK was extremely pleased with the performance improvements in DB2 Sort Version 2.1 over DB2 Sort Version 1.3. DB2 Sort Version 2.1 delivered multi-record processing to further reduce the overhead between DB2 Sort and DB2 Utilities and tools that call DB2 Sort. This feature results in even greater processor utilization reductions than ever before and helped increase zIIP offload. BITMARCK also experienced more than 10 percent savings across all areas when running the Version 2.1 beta. BITMARCK also reported the following observations: 

  • DB2 Sort Version 2.1 is a very stable, easy-to-install application that is now in multiple environments, such as production and quality assurance (QA). There are absolutely no changes required to any applications.
  • The DB2 Version 2.1 batch window was shorter, and it performed faster than the batch window and speed, respectively, of DB2 Sort Version 1.3.

Achieving sort optimization

Data is the newest resource for competitive advantage. DB2 Sort is well suited for organizations that have DB2 Utilities processing increasing amounts of data with reduced processor usage, improved elapsed time and increased zIIP processing. DB2 Sort requires no change to any business-critical applications and helps meet demands for data growth across z Systems.

Maggie Li, coauthor, is a manager, mainframe development, at Syncsort Incorporated.