May, 2008
by Denise P. Kalm, Secretary, NCCMG
Denise Kalm has 30 years experience in IT including application programming, enterprise systems management and performance management/capacity planning at Pacific Telephone and Bank of America. She moved to vendor land in 2000, spending 5 ½ years with BMC on the EPA product line, then recently became the senior product marketing manager for enterprise job scheduling products at CA, Inc., formerly Cybermation. She is a regional officer of CMG, has held many volunteer positions within that organization and is a frequent contributing author. Prior to entering the IT profession, she was a biochemical geneticist. Her hobbies include flying, Jazzercise, writing and scuba diving. Her book, Lifestorm, on the Oakland Hills fire, is available on Amazon. She is an executive and personal coach as well, offering phone and in-person coaching.
About the Author
Denise P. Kalm, CA, Inc. formerly Cybermation
Our President, Cathy Nolan, found us a great new location for meetings - the Carr America Conference Center in Pleasanton. The location appears to be a hit with all, as we had a very large attendance, including some new and some long-time absent old faces. Sysload Software Corporation kindly hosted our meeting providing a yummy breakfast and a great sandwich lunch buffet. Sysload introduced their product line to our membership at CMG2007 as a gold sponsor and indicated a desire to re-up for 2008.
We began the day in real time with a talk on virtualization. One of the Sysload team, Emmannuel Sauvion, presented "Performance Management in the VMware World." Mr. Sauvion has been with Sysload for 6 years, but is a newcomer to the Bay Area, having moved just 4 months ago from France. He began by giving us an overview of virtualization, describing the container technique used on the Sun Solaris 10 and others and the partitioning technique exploited by AIX, VMware and others. Because of the proliferation of guests and the distribution of resources, he suggested that random saturation at the guest level can be a real problem, as in the hypervisor case, the switch decision is based on averages. He then spoke of the layers of management needed, noting that the view of CPU demand differs between the host and the guest, and must be reconciled to have a true picture of demand. Sysload believes that you need to collect metrics at the 5 second granularity to reflect micro-climate changes and saturation. He then introduced how Sysload collects data and helps you manage this challenging environment with little overhead.
Nalini Elkins, CEO of Inside Products, Inc., presented her new, patented method for analyzing packet traces, "Visual Diagnostic Language (VDL) for TCP/IP Trace Analysis." She began by noting how hard it is to actually read traces, even for the experienced network administrator. She felt it could be easier. Knowing that our brains actually do a great job of pattern recognition, if we have a visual image to work with, she created a tool that transforms a lot of packet text into a visual representation of packet flow, including source and destination, packet size, direction, congestion window and response time. Ms. Elkins noted that written text requires use of our frontal cortex, a part of the brain developed long after the part that looks at patterns. Small windows indicate congestion; large hour glasses indicate bad response time. She also includes narrative optionally to explain the data and binoculars if you wish to drill deeper into the data. There is also the option to filter the trace as you want, to focus on the elements you care about. A new capability is "sonification" where the patterns can be heard, just like a Geiger counter. For many people, they can actually pick up the patterns faster by ear than by sight. She noted that Alice Walker "hears" dialogue for her books, then writes it down. Ms. Elkins finished with a demo.
We had some time to network over lunch and then Sysload returned to show us how their product worked. We got a number of handouts to take home and peruse and at the end of the day, also got the first Sysload shirts on the West Coast - a style statement if there ever was one. It should be noted that their patented data collection methodology does feed into all the major monitors, so install bases can benefit from this capability without making a huge amount of change or a steep learning curve.
Jon Schmidt, Transaction Design, Inc., spoke to us about "Managing the Customer Experience." He described the change in our world that insists we manage proactively; we cannot afford to wait for user complaints because the cost is just too high. He began by defining what an SLA can be (availability, recoverability, throughput, response time or a combination) and noted how few people have really good SLAs in place. Using examples from real customers, like banks and retail systems, he showed his methodology, which involves recording information about every transaction so you can easily find the cause of a slowdown - the important information is in the outliers and the patterns. Mr. Schmidt likes to look especially closely at peak ½ hour data over time to note trends and degradation. Year to year comparisons can show business cycle patterns, which can help you predict future demand. He suggests reporting on response time in multiple buckets, such as ideal, OK, on the verge and bad. This helps you set up early warnings. The examples made it clear how much value this method has, and we really already have the data - we just need to consider looking at it in this way.
Tom Halinski, Compuware, joined as again (he has a preference for NCCMG gigs in November and May to escape Michigan weather). He talked about "Business Service Management," which actually turned out to be a very good linking together of all the preceding presentations. First, he defined the age-old problem - relating business to IT and vice versa, but he showed ways to bridge that gap. As he has in the past, he referenced Apdex as a valuable metric to communicate between both sides and the importance of getting away from silo management, as it misses the connections between the silos, often resulting in a "no problem found" situation. To Mr. Halinski, the "secret sauce" for BSM is the service model - a way of representing business and IT services and the way they connect, so that reporting speaks to both the IT demands and the business importance.
Our next meeting is August, 2008, probably at the same location. Save the date! Any questions or comments, please contact Cathy Nolan at ( or ).