This year at VMworld (both San Francisco and Barcelona) performance will be front and center. I've been working internally to create a "mini-track" of technically advanced performance breakouts with many of our actual performance engineers as speakers. Customers always want to know about best practices, troubleshooting and just how far vSphere can push the performance envelope and that's very hard to do in a 60 minute session. So this year I've got approval to try something different.
There will be a four part "Extreme Performance Series" set of breakouts that will allow the speakers to dive deep into their respective areas without overlapping of content. This means you get to spend a whole breakout on an individual resource dimension: CPU/Memory Scheduling, Network, Storage and vCenter. Dive deep into the architecture, new capabilities, recommend practices and troubleshooting of each resource dimension. You'll want to attend the whole series in order to learn about the complete stack.
I've created a blog post to highlight each session and why I'm excited about it.
VSVC4811 - Extreme Performance Series: Monster Virtual Machines
Seongbeom Kim, Performance Engineering, VMware
Peter Boone, GSS, VMware
Mission critical applications serving an entire organization represent the last hurdle to the organization looking to migrate its data center to a private cloud. Although, IT administrators realize the benefits of virtualization, they may be hesitant to virtualize the mission critical applications because of the resource demands of these applications and an inexplicable fear about the underlying hypervisor preventing the infrastructure from meeting the SLAs of these applications. vSphere, industry's leading hypervisor platform, has been pushing the boundaries of infrastructure resources it can manage over various releases. Through its superior resource management capabilities, vSphere can easily scale horizontally to support many virtual machines (VMs) or vertically to support large VMs. Recently, vSphere extended the resource management envelope to such a level that even extreme resource-hungry applications such as databases can be run in VMs. In this talk, the speakers will give an overview of key vSphere features such as CPU and NUMA schedulers, and memory managers that propel monster VMs. The speakers will share their experimental results, experience gained, and lessons learned. Finally, the speakers will provide best practices to the audience for undertaking such an exercise. This talk should alleviate any hesitation to migrate the resource hungry, mission critical applications to a private cloud.
I can think of no better speakers than Seongbeom and Peter for this session. Seongbeom's daily efforts are about diagnosing and improving the performance of these resource dimensions. Peter, an experienced performance guru in our support organization, and fellow Canadian, spends his days working on the front line with customer issues. While the abstract above references Monster VMs a number of times, all the information in this breakout can be applied to environments of any size. They explore processor technologies, how our schedulers work, NUMA and its performance implications, identifying and fixing scheduling contention, etc. so that next time you hear "my VM is running slow" you'll have experience, tools and confidence to prove otherwise.
Excited to see you all there!
See the current VMworld 2013 content catalog here: https://vmworld2013.activeevents.com/connect/search.ww