October, 2007
by Linwood Merritt
Lin started his data processing career in 1970 as a Simulation Analyst, has been a US CMG member since CMG84, and has been the Project Manager of the "Enterprise Wide Capacity and Performance" project of SHARE since August 1994.
Lin has published 14 CMG papers (not counting 1 additional for UKCMG 2003) and presented at 25 SHARE conferences. He won the CMG97 Mullen award for the presentation "Performance Data from the Server to the Intranet: Getting the Data and Reporting It," and presented it at UKCMG in 1998. He is now working as a mainframe Capacity Planner at Bank of America in Richmond, Virginia.
About the Author
Linwood Merritt, Bank of America
Southern CMG held back-to-back meetings in Richmond, VA, on September 27 and Raleigh, NC, on September 28. The logistics of consecutive days allowed the region to double-book Rick Ralston (presenting "Using SAS/Graph to Display Performance and Capacity Data") and Frank Bereznay (presenting "Using Statistical Techniques to Interpret Service and Resource Metrics"). The region also double-booked breakfast sponsor MVS Solutions Inc. and lunch sponsor BMC Software.
The Richmond meeting was kicked off by Igor Trubin's "System Management by Exception," his "final" presentation in a series of papers describing techniques to identify exceptions for a large number of computer systems by applying statistical methods. Igor's approach is based on Multivariate Adaptive Statistical Filtering (MASF), and tracks hour-of-day and day-of-week historical data to identify hourly data points above or below three standard deviations.
Rick Ralston's SAS/Graph presentation (at both meetings) described the use of ODS (Output Delivery System) to enhance graphical output, discussed coding tips and described the production of "heat charts" (similar to temperature charts in weather maps) to present a large amount of three dimensional data in two dimensions. Examples included system utilization or performance by hour-of-day for an entire month.
Kathy Walsh presented "z/OS Hot Topics" in Richmond, an update to Washington System Center issues and recommendations. The topics included z/OS timing changes, enhancements to the zPCR mainframe sizing tool, new encryption functionality in the zIIP specialty processor, and LPAR Groups (defined capacity limits for groups of partitions to limit software billing).
Frank Bereznay presented his Mullen Award paper on statistical techniques in both meetings, describing four approaches: Hypothesis Testing, Statistical Process Control, MultiVariate Adaptive Statistical Filtering (MASF), and Analysis of Variance (ANOVA). One important point that Frank made was that the first two techniques are based on random sampling (not generally found in fixed-interval computer data). MASF, developed by Annie Shum and Jeff Buzen and presented at CMG95, is a statistical detection technique based on interval driven time series, overcoming the "random sampling" requirement. ANOVA was presented as an excellent complement to MASF to identify statistically-similar intervals to combine for analysis.
Peter Sevcik's "Benchmarking APM Best Practices," presented at both meetings, discussed the management of user flows in Application Performance Management and compared the results of user site surveys with responses evaluated on a scale on 0 to 10. Attendees of SCMG, given an opportunity to respond to the survey prior to the meeting via an online link, scored slightly higher (5.2) than the average score of 4.3, and was within the top third of responding sites. The results were broken down by four distinct APM process stages (Understand, Measure, Communicate, and Link).
"A Queue Simulation Tool for a High Performance Scientific Computing Center" was presented by Jim McGalliard in Richmond. Jim described a tool written in C that simulates job scheduling options in a massively parallel computational batch environment. The tool evaluates impacts on schedules when jobs are moved around in the schedule to backfill idle capacity openings in a nightly batch schedule.
Chris Molloy presented "The "Powerful" Capacity Manager" in Raleigh, discussing Facility Capacity Planning, a capacity planning discipline that has been added to ITIL V3. A major component in this arena is power consumption, especially in today's push for "green" environments (with a green grid component that takes the ratio of "power coming in" to "power used by IT equipment"). His presentation included a discussion of the evolution of datacenter efficiency from centralized, to physical consolidation, to virtualization, and to application integration.
"The Myth of MSU" was presented by Jim Horne in Raleigh. Jim discussed mainframe measurements of general purpose processors and newer specialty engines (zAAPs and zIIPs), concentrating on two levels on system accounting records (Type 70 and Type 72). The use of Service Units in reporting and analysis was discussed (and was a focus in the title), particularly when dealing with the new specialty engines and the different ways that they are reported.
The Raleigh meeting was concluded with "Virtualized Capacity - Myth or Reality?" by Susan Schreitmueller. Susan discussed virtualization on IBM Power systems (midrange UNIX).
The Southern Computer Measurement will hold its next meetings in the spring of 2008. Anyone interested in being a sponsor or presenter or who has any suggestions should contact .