John Troyer offered to take us on a tour of the VMworld Lab facilities on Sunday. It was great to get a preview of the Labs, but what was more interesting to me was the underlying technology that VMware has implemented to make this happen.
Fun Lab Facts:
- 480 Concurrent Sessions
- 150 Subject Matter Experts assisting with the lab
- 8-11 VMs in a Lab Pod (the environment you connect to and interface with)
- Pods consume up to 27GB of RAM each (so called Sandbox aka “Kitchen Sink” lab)
- 100′s of TB of EMC & NetApp storage at each site
- HP & Cisco blades for compute
- 12 Lab Manager instances running in parallel, controlled by “Lab Cloud” custom integration
The Facilities
All the labs are in Moscone West this year in one large room. Here’s a shot of the basic layout:
Figure 1: VMworld Labs
The setup includes a Wyse thin client with two displays. The idea here is that one display shows the lab manual and the other is the remote session used to access the labs.
You can see the GbE cable bundles at the end of the table. These are aggregated in central areas on stacks of Cisco 3750s.
Figure 2: GbE aggregation points
The GbE links are aggregated to 10G and connected to Cisco Nexus 5k Switches in the DC area, which we didn’t get to see.
Hybrid Lab Systems
Another interesting and cutting edge component of the labs is the use of multiple redundant remote datacenters to provide some of the capacity. There is a third of the physical capacity onsite, with another third in Ashburn, VA, and the final third in Miami, FL. There are two 45Mb pipes to these remote DCs.
This provides redundancy in case there are issues with any of these datacenters. It also reduces the overhead for setup for Copenhagen as the plan is to swing this capacity across transatlantic links to drive the labs there. This is a very clever design that gives the Labs great resiliency and redundancy.
Content Delivery
Here’s a shot of the basic setup with the thin client and displays.
Figure 3: Thin Client setup
On arriving at the lab you will need a proctor to log you in. Once logged in you’ll see the following:
Figure 4: Lab Track selection
The screen on the right allows you to select the Lab Track you want to pursue. This is necessary as there are so many Lab Sessions there needs to be an extra interface at this level or things get cluttered. As it is, there are still plenty of choices for each track.
Figure 5: Technology & Architecture lab offerins
Here are the labs offered in the Technology and Architecture track. Once a session is selected it will either connect to a prepopulated lab environment or the environment will be loaded on demand. This is one of the tunables that the Lab staff will be working to balance over time. The more popular sessions will have prepop pools that are available. The rest will load on demand in 5-7 minutes, so be prepared for a delay for some of the labs.
Figure 6: VMworld 2010 Lab Manual interface
Here you see the lab session on the left and the manual screen on the right. There’s a countdown timer that shows how much time you have left to finish your session. If the labs are not full you can add time. This is another tunable that the staff will use to manage traffic. There’s also a Get Help button which will get the staffs’ attention so someone can come by and assist you.
This is an amazing piece of technology on display. Remember, the user with the most completed labs will win a pass for VMworld 2011. Please take advantage of this opportunity and stop by the labs early and often.
August 30th, 2010
Posted by
Jae Ellers |
Uncategorized |
2 comments
Bryan Semple, VKernel CMO, spent some time with me last week reviewing some of the new features and improvements in their Capacity Management Suite (CMS). According to Bryan, there are three improvements users should hear about in the new release. These are:
- Scalability
- Analytics
- Automation
We’ll review the improvements in these areas in more detail below
Here’s a shot of the dashboard:
—
Figure 1: CMS 2.0 Dashboard
The VKernel CMS is composed of four components that can be mixed and matched. Together these form the entire suite. If you’re not already familiar with the various components please visit the VKernel site and take a look at the full feature list, as I’m just highlighting some of the new functionality. These components are listed below with some of the recent improvements:
- Capacity Analyzer 5.0: Nuanced analysis allows fine tuning of what resources are reported on and which systems to exclude
- Optimization Pack 2.0: Automation and scheduling enable fine grained and extensive reclamation and recovery activities
- Chargeback 2.0: Report on allocated or utilized resources for VMs
- Inventory 1.0: Inventory, index, filter, document, annotate the environment and configuration
One of the additional improvements we discussed was streamlined installation. This comes from the fact that the entire suite is now integrated into a single VM with licensing activating the various modules. This helps improve the performance, scalability, and manageability of the product by reducing the footprint and inter-component communication necessary.
VKernel has also been working hard to improve the UI. When dealing with 100s or 1000s of VMs the quality of the UI has a huge impact on the utility of the product. From what I’ve seen, the new interface reduces time and clicks when reviewing and optimizing your environment using VKernel’s custom groups. When this is coupled with the ability to exclude vms, hosts, clusters, and run analyses only over custom time ranges it provides a new level of accuracy in reporting.
One great example of this is that most VMware admins have a test cluster or a dev environment (or both if you’re lucky). We don’t want those older GHz which are frequently idle to count against us for production machines when reporting to the CIO. With the VKernel suite it is a simple matter to exclude these systems from the analysis and reports.
All together, these improvements add up to an increased ability to monitor the resource utilization as you wish to see it while concurrently capturing configuration data. This is important as you have correlated historical configuration data to review as you tune your environment. This can be a real lifesaver should you need to back out config changes in your environment at a future date.
Finally, the real killer feature here is the fine grained control and scheduling ability to recover and reclaim wasted resources. This really leads to a more self-tuning environment where the automation delivers actual ROI by providing more available resources for more VMs or enhanced capacity for existing systems.
August 26th, 2010
Posted by
Jae Ellers |
3rd-Party, Uncategorized, integration, management |
2 comments
Hola Amigos,
It’s been a long time since I rapped at you, but things have been crazy. Changed jobs, remodeled a couple houses, had a baby, etc. I hope to get back to normal soon.
I have a couple things on my todo list for this blog (not surprisingly), the first of which is to put out a new VC2.5U4 visio schema.
I have also reviewed Hal’s upcoming book on the VIToolkit and Powershell: http://halr9000.com/article/660.
Finally, I’ve been working a lot with ESX, SRM, and vSphere. I’ll still try to split up the posts with more of a market and architectural spin here with tech details on vmprofessional.
More RSN.
March 10th, 2009
Posted by
Jae Ellers |
Uncategorized |
2 comments
