November, 2008
by Tom Confrey
The Connecticut Computer Measurement Group (CCMG) held its Autumn Conference on October 17th. It was, again, a mix of local, national, and international speakers. There were two parallel tracks in the morning followed by a single track in the afternoon. Not only did attendees expand their knowledge and share insights at this informative meeting, but a beautiful fall foliage arrived and presented to all those who ventured outside.
The first presentation in the morning's Technology Track was by Tony Mungal who spoke on Multi-Core Processing and the Tiered Storage Model. Tony is a Consulting Corporate Systems Engineer with the EMC Corporation. He is based out of Boca Raton, Florida. He has authored and published numerous papers and has presented at forums such as Computer Measurement Group (Regional, National, and International meetings), SHARE, GUIDE and an assortment of other IT related user groups. Tony is also President of CMG Canada which only adds to his reputation as a man of international intrigue.
As you know, most enterprise class I/O subsystems today employ multiple types of storage technologies meant for deployment using a "tiered storage" model in support of the trend of increasing capacity and performance driven largely by application and computing trends. Indeed, tiered storage as an effective I/O strategy is a necessary step in the ongoing effort to capitalize on technological developments while lowering overall cost to the user. However, there is a tendency for these deployment strategies for application and consolidated workloads to become convoluted, and even ineffective.
Tony's presentation discussed the multi-core and multi-processing computing trends which affect application deployments, and its pronounced effect on the I/O subsystem. He covered topics such as data growth, access patterns, traffic characterization, and the working sets that were used to introduce the need for technologies such as larger drives, flash disks, multi-tiering, dynamic cache partitioning, virtual provisioning and the like. (Yes, multi-tiering is a new word.) Tony further examined the technologies, their lifetime expectancies, their suitability to various workloads and applications, and where applicable, the next steps. A most informative and interesting session. The second session in this track was given by Irwin Kraus, who is well known to many of us in CMG and who is the North East Technical Sales Manager for Verari Systems. He has 40 years of experience in the IT industry, both as a customer and a vendor, and has been a frequent speaker at CMG and SHARE conferences in the United States and Europe. Irwin is a former mainframe guy and the name of his presentation was x86 Grows Up, which basically was how x86 clusters are an alternative to mainframe and mid-range platforms. (Note, we said, "former" mainframe guy.)
Irwin shared the view that there is a gradual shift underway in enterprise computing from mainframe and midrange systems increasingly toward x86-based systems. This shift is being driven by several factors including the high cost of mainframe and midrange hardware and software, and is largely enabled by the growing capabilities and performance of x86 based systems. Other major factors include strong open source OS and middleware as well as robust virtualization offerings. Finally, Irwin contends that a new stateless, highly parallel application architectures are emerging which are ideal for large clusters of x86 systems. His presentation was very informative to those trying to understand where x86 is being positioned in large enterprises in relation to the IBM mainframe, along with the trends of the surrounding technology such as network, storage, OS, virtualization, applications, availability, etc. It was a very interesting and helpful session.
Running in parallel with the Technology Track was the Open Systems Track. The first speaker was Eric Landrieu, who recently presented at Philadelphia CMG. He discussed Managing Capacity and Performance of a Virtualized Data Center. Eric has worked for PERFMAN since 1997 and is a Software Development Manager responsible for PERFMAN for VMware. He is also a VMware Certified Professional.
In his session he discussed important metrics and best practices in managing virtualized data centers. He also explored the possible consequences of improperly configured servers and virtual machines. He covered various virtualization technologies, with VMware as the primary focus. Eric was quite knowledgeable on virtualization and examined real life issues and provided solutions. His session was very much appreciated by attendees.
Following Eric in the Open System Track was Debbie Sheetz, who gave a presentation entitled Predicting the Relative Performance of CPU. She is a Capacity Management Architect and Support Analyst based in BMC Customer Support. Originally hired to work on the first version of BEST/1 at BGS Systems, she has many years of experience developing and supporting performance analysis software.
When an analytic model or sizing tool predicts CPU Performance, it uses (1) queuing theory calculations and (2) a representation of relative performance for each CPU configuration. Debbie's session focused on the second part (representation of relative performance for each CPU configuration) by using a sample CPU rating system from www.spec.org. Her research results include an algorithm for finding functions to describe relative performance, as well as specific properties of general and vendor-specific functions. Deb showed how these results can be directly applied to any case where the CPU vendor supplies limited benchmark data compared to the number of CPU configurations actually available.
After lunch the groups joined together for the afternoon Combined Sessions. First up was Ron Potter, a CCMG member and friend, who gave a presentation on the best practices for Writing a Capacity Planning Report. Ron has been involved in IT for over 30 years, working in a number of different industries. Before "retiring", he was closely involved with a multi-year IT improvement initiative at a healthcare company that embraced the ITIL framework and associated best practices. (We put retiring in quotes because, as far as we could tell, Ron is still doing a lot of IT work.)
If you have never written a capacity planning report or not quite sure what should go into one then, obviously, this would be a session for you. However, because this type of report is one of the most important aspects of the capacity planning process we felt it might be a good idea to listen to what Ron had to say. (We did listen and we did learn.) He noted that the report should clearly explain the results of your work, help management understand any issues, and provide the information necessary to drive consensus on getting any actionable items resolved. Ron also provided ideas on simple ways to use PowerPoint slides to tell your story with impact that technical, IT management and business staff will understand.
Yiping Ding, who has presented many times at CCMG (some claim once a year, but he claims every couple of years) as well as national and international CMG conferences, completed our day with a presentation on Analytic Modeling for Application Performance Index. Yiping is a Vice President of R&D for OPNET Technologies. He is also responsible for Panorama and Systems Modeling Solutions at OPNET. Before joining OPNET, Yiping was a Chief Scientist at Akorri. (And, oh yes, he also worked for BGS Systems and BMC Software for more than 15 years as a scientist, senior manager and architect.)
This session was quite interesting because it wasn't that long ago that some of us dreamed of reducing all of our performance data down to one number per day to reflect how things were. For those of you who are not familiar, Application Performance Index (Apdex) is a new standard in the characterization of response time based on user satisfaction. (Check it out at www.apdex.org ) It is a single number used to quantify user satisfaction with application performance. Yiping has written a paper that presents an analytical model to study the relationship between Apdex and the capacity of the system that supports the applications. The analytical model provides system-capacity guidelines to achieve the desired Apdex rating. He also explained how it helps one to better understand the correlation among application response time thresholds set by the users, the user satisfaction level, the Apdex value, and the mean response time delivered by the system. A very good session and a step closer to that one number.
We thank TeamQuest for sponsoring the Continental Breakfast and Refreshment Breaks. Their sponsorship helps subsidize the price of admission to our CCMG events.
Our next meeting is scheduled for Friday, April 3, 2009.
Please visit us at our web site for more details on past, present, and future meetings. regions.cmg.org/regions/ctcmg/