Engineer Mark Manness’ breakout session focused on understanding that performance instrumentation at all levels (e.g., application, guest OS, virtual machine, host) is necessary for successful capacity management on VMware. The three main takeaways from his presentation:
- Monitoring inside the Guest OS is an important piece of the performance puzzle
- What some of the Host and Virtual Machine metrics mean and how they can be used.
- Appreciate the numerous modeling options available on the VMware platform.
- Physical to Virtual planning
- Proper configuration of virtual machines
- Using data from smaller benchmarks to size larger workloads.
Thanks for the post Mark.
TeamQuest engineer Scott Johnson wanted the audience to walk away with the following main points from his presentation:
- While measuring application response time can be a challenge, there are some reasonable options for passive data collection.
- A fortuitous side effect of some response time monitoring agents is additional data that can be used for application trouble shooting and business intelligence.
- Application measurements can be used to enhance capacity models.
Johnson also noted that transaction rate and response time information are very valuable for monitoring service level agreements. In the world of Unix, Linux, and Windows, there is no standard transaction monitor for applications. For those that grew up in the mainframe world with proprietary transaction monitors, this situation is a bit unsettling. However, several approaches have been developed to address certain classes of applications.One approach for web-based applications is to use synthetic transactions and response time “robots” to initiate and measure these transactions. This approach is called active measurement since it initiates a small artificial load on the application in order to obtain measurement data. This approach provides reasonable indicators of application availability and network latency.Another approach is to instrument the application and then make the response time information available, perhaps through an application log file. Also, the technology used in web applications provides the possibility for web server log file analysis and TCP packet sniffing. These last approaches are usually called passive monitoring since they add no transaction load to the application. The advantage of passive monitoring is the potential to understand the actual rate of application activity and the actual user experience.Thanks to Scott for this post.
The guy pictured above was a part of the entertainment at the North American TTS 2010, where this year’s theme was “Optimize Your IT Zoo.” The San Diego Zoo is famous throughout the world, so in addition to the technical content, there were a number of visitors present who work at the zoo, including the beautiful bird shown here.
In some cases the technical content of the show was blended together with the zoo-based theme. In fact, one of my favorite presentations at this year’s event was a talk by the zoo’s Chief Technical Officer, Robert Erhardt. Here are a few tidbits from my notes:
- Google Street View works inside the San Diego Zoo. Check it out!
- The zoo attaches tiny fraction-of-a-gram RFID devices with little whisker-like antennas to rare mice in order to track them.
- Condors have radios that send data via satellites. The trick needed to make the radios practical was to design antennae that did not interfere with the birds’ ability to fold their wings.
- In California, organizations can qualify for big rebates by implementing power-saving virtualization solutions. They measure your power consumption before and after virtualization and your rebate is calculated based on the reduction in power usage.
I thought Erhardt’s talk was both entertaining and informative. It was good fun that I think exemplifies this year’s summit. If you didn’t make it this year, don’t miss next year’s event which is sure to be even better. At next year’s summit we will celebrate TeamQuest’s 20th year of existence!