On April 22, 2009, we had a great conference that included presentations from two invited speakers. IBM Tivoli was our sponsor. 

In the morning, Peter Sevcik, the inventor of the Apdex score spoke. Peter also the executive director of the Apdex Alliance and a principal in the NetForecast website. He is a leading authority on measuring, assessing, and improving the performance of networked applications. He has contributed to the design of more than 100 systems and to the success of more than 25 application management products. He has pioneered several performance management techniques that are in common use today.

The Apdex score is the open standard numerical measure of application performance satisfaction. Essentially, the Apdex score is a number between 0 (Zero) and 1 (One). Zero represents an application performace that is rated by the user community as being a disaster and one represents application performance perfection. The base idea of the scoring system is to achieve a rating, such as point five nine's as defined by the user community. A user score example would be described as being an "F" (frustrated with the application) ,  "T" (tolerating the performance),  or "S" (satisfied with the performance) of the application. The strength of the Apdex scoring system is that it discovers hidden problems. The use is not 'one time', but is a periodic trend that can be used to help find capacity problems during the time addressed. The important part of the system is that the score is linked to the business needs. John showed the group his 'formula' and used a case study of a business-to-business commercial website as an example of use. He explained some of the different metrics selected as the measurement markers.  The Apdex score is used by many large companies and government web sites.

Peter's presentation may be viewed at http://www.apdex.org. He suggested further study at the following site: http://kpilibrary.com.   (This contains Key Performance Indicators used by the scoring system.)  Also, others may be interested in the web-site http://www.netforecast.com where Peter is a frequent contributor to the blog area.

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

In the afternoon, our second presenter was John Van Wagenen. John works with a team of performance analysts at Caterpillar, Inc.  The company is a large producer of diesel engines that includes the ones used in ships, trains, and trucks. He attended national CMG for the first time in 2006 and presented his first paper in 2007. Following that experience, he decided to write three papers, in hopes of having another accepted. Well, all three were accepted and at the national CMG 2008 conference, he presented his papers and was selected as a Mullen Award winner!  With this award, comes a stipend that allows him to travel to regional meetings and present his papers to others. He is an engaging speaker and is truly interested in helping others learn from his experience.  We are grateful to John for coming to visit us. John didn't spend much time in Atlanta, he flew to Atlanta the morning of the conference, and he flew out immediately after his presentation. He told us that he is also presenting his paper in Paris, and other regions in the upcoming months. 

John presented two papers. His first paper was: “So you want to manage your z-Series MIPS? Then detect and control application workload variance!”  He has an outstanding success story from Caterpillar where they have not upgraded the mainframe capacity for two years, in spite of  doubling growth in sales. He explained how his team was able to find code inefficiencies by analyzing usage and comparing the baseline measure of an application to recent growth, or variance. They look for applications that exceed a 20% monthly threshold and target that application for more analysis. Many times, a deeper dive into application performance yields finding code that can use less computer cycles if it is changed to be more efficient. His paper was the result of lessons-learned and instructions on how others may measure their usage. He creates a daily/hourly file for about 700 applications that run on the mainframe. He noted that they execute 10,000 batch jobs an hour. The team looks at schedules that can shift and many factors of how the mainframe can be more effective. 

John's second paper was titled: “Pivot tables/charts - magic beans without living in a fairy tale".  This presentation was one where ‘you had to be there’. John was both entertaining and educational with a live demonstration of the creation and use of Microsoft Excel Pivot tables. The take-home use of this information is invaluable to those who have not experienced ‘pivoting’.  He has found where delineation of functionality creates better diagnostics. He used a DB2 Authid/CorrelationId pivot chart as an example for the web activity at Caterpillar. He said that they have found that sizing workloads can improve application forecasts. They use a technique called "Pad Management" where they follow application growth patterns, and tune, when possible, to protect the white-space. (The area on a graph above actual usage and below the maximum capacity of the machine.)