performance tuning service for IBM Mainframe Antonio Jose

performance tuning service for IBM Mainframe Antonio Jose Ferreira

Reducing operational costs through MIPS management Information Technology plays an important role in many organizations IT related operational costs are high and no wonder that businesses ought to strive for their reduction For most banks, it was estimated that total operational IT costs oscillate at around 15% to 20% of total operational costs Businesses have taken a strong stance on operational costs reduction and sought solutions to slash IT costs For instance, Citigroup estimated that removal of redundant systems from their IT portfolio yields a savings potential of over 1 billion USD

Reducing operational costs through MIPS management So far, organizations have tried a number of approaches to cut IT costs which included head count reduction, outsourcing, consolidating data centers or replacing old software and hardware with its newer counterparties All these approaches, however, carry an element or risk and might end-up costing a lot especially when complex software environments are involved There exists one approach which enables IT costs reduction at low-risk for both business and IT By enabling control of MIPS consumption not only it is possible to achieve significant savings but also to improve performance of IT-systems

Reducing operational costs through MIPS management MIPS have envolved to become a measurement of processing power and CPU consumption MIPS related costs are inevitable and high for those organizations which have grown on top of mainframes A former CIO of a corporation, once said: “MIPS and salaries are my highest unit cost per month” The amount of MIPS used by the average IT organization is on the rise IT analysts estimate that most large organizations utilizing mainframes should expect their systems’ MIPS consumption to increase by 15 -20 per cent annually

Reducing operational costs through MIPS management Each additional MIPS typically costs around GBP 2, 500 in hardware and software charges So if a company running a 10, 000 MIPS system increases capacity by as little as ten per cent per annum, the incremental cost will be in the region of GBP 2. 5 million Despite the fact that the incurred MIPS related costs are substantial the majority of IT executives (60 %) do not monitor MIPS usage Therefore, reducing MIPS is indispensable and is one of the major priorities of mainframe users, due mainly to a weak economy and high software costs on the platform

Reducing operational costs through MIPS management Where does the accrual of MIPS happen? MIPS consumption is directly driven by CPU usage, and CPU usage depends on the applications’ code Inefficient code of the applications is considered to be one major driver for MIPS usage According to a financial services company identified two lines of code that, once changed, saved 160 MIPS Given the market price of MIPS, the two lines of code contributed to substantial reduction of operational IT costs

performance services in a z/OS 25 WORLD Reducing MIPS growth is a top priority 20 15 10 Ideal MIPS usage Why pay more for inneficient code? 5 The 0 worldwide installed base of IBM mainframe exceeded 14 Million MIPS by end of 2008 - up approximately 20 % from year-end 2007 and roughly double of what was five years ago. The need for controlling MIPS attributed costs is indispensable given the high percentage they involve in the overall IT costs.

recommending a strategy Code optimization COBOL · SQL · JCL Reduction MIPS & I/O usage BATCH window reduction

recommending a strategy Our approach relies on the mainframe usage data, facts extracted from source code, and is supported by a real MIPS-reduction project which involved tuning of COBOL Implementing standards-oriented SQL JCL programming performance that allows: Increasing performance that results in MIPS reduction More workload to be processed without adding additional resources Allowing for growth

recommending a strategy Efficient COBOL Coding Techniques Specific rules for performance, typically oriented to the code level, may lead to significant improvement in performance, and should be a good starting point. More than 150 rules are available to achieve a better performance. JCL standards and efficient use The bad choices in JCL programming, may have a very negative effect on system performance, specially in cases involving a very large amount of information. Increasing ‘parallelism’ and an intensive use of IBM utilities. SQL performance and tuning The vast majority of the top MIPS consuming operations in the production environment involve the use of DB 2. MIPS reduction at low-risk and also low-cost by improving SQL code on the basis of mainframe usage. More than 200 rules and tips to improve DB 2 performance.

SUCCESSFUL REAL CASE Sample Tuned Charts ODATE OLD NEW REDUCTION Day 1 3, 28 1, 26 Day 2 3, 29 1, 17 - 61, 59 % - 64, 44 % Day 3 3, 46 1, 28 Day 4 3, 46 1, 22 - 63, 01 % - 64, 74 %

SUCCESSFUL REAL CASE Sample Tuned Charts ODATE OLD NEW REDUCTION Day 1 73, 91 32, 32 Day 2 73, 99 20, 74 - 56, 28 % - 71, 97 % Day 3 91, 82 32, 90 Day 4 82, 22 20, 77 - 64, 17 % - 74, 74 %

Conclusion Recommendation THE CHANCE TO ACHIEVE THE BEST PERFORMANCE ü Improving individual program performance ü Improving the parameters of the data sets employed ü Increasing parallelism (through several methods) ü Splitting or merging programs GOAL Improve performance Reduce CPU usage Reduce costs ideas for today and tomorrow