Xangati: A Combination Of Bacon, Star Wars, and Performance Monitoring
At Gestalt IT’s Virtualization Field Day 2, delegates should expect the unexpected. No one seemed to understand that better than Xangati, who was a sponsor of the event and gave a presentation on their virtual performance monitoring products. In tow, they had nearly the entire cast of the Empire Strikes Back set up in cardboard cutouts with various “blame” slogans emblazened upon them, along with all sorts of bacon and chocolate treats. This was amazingly over the top and was a huge eye popper. Being not that familiar with the Xangati product, I was worried it may have been a play to cover up a lack luster product. Fortunately, that’s not true, and I’ll use this post to go over some features that I found interesting and noteworthy.
Some familiar faces spotted at the Xangati presentation
The Xangati product uses a management VM, labeled the Xangati Management Dashboard (XMD), along with per-host service VMs, called the Xangati Flow Summarizer (XFS), for data gathering. The XFS VM must be deployed on a portgroup set to promiscuous mode (allows sniffing of all traffic). Anyone who’s worked with virtual ESXi hosts should be familiar with this process: it’s just a checkbox and can be done on a live portgroup. Optionally, if you’re using vSphere Distributed Switches on vSphere 5 (requires Enterprise Plus licensing) you can take advantage of the NetFlow exports instead of XFS VMs. I’d imagine that you’d want to play with the sampling rate if your vDS is pumping a lot of traffic through the host, as it could potentially cause a performance bottleneck. Tomi Hakala at vReality has a rather clean write-up on setting up NetFlow on a vDS for further reading.
A snapshot of the dashboard while at Virtualization Field Day
In order to get going with collectors, the XMD (Dashboard) has a initial setup wizard that prunes networks you want to monitor, and also pushes out XFS service VMs to your hosts. It is a bit “click intensive” to go through each host and select both the portgroup for the PCAP interface and deployment datastore. I’d like to see a way to have this monitored proactively as hosts are added to the system. However, the process is not difficult by any stretch of the word, and setup time for me was roughly 30 minutes in a lab environment (including watching the install video by Nathanael Iversen, the “video voice” of Xangati).
They are strongly appealing to my geeky side. Also, why am I holding my hand like that?
In regards to data gathering, I’ll give the product a gold star for not simply using the vCenter statistics database, as I’ve found this data to be both unreliable and far too averaged for my taste. It’s hard to track a spike with so many rollup averages being performed, and attaching to a NetFlow stream is a great way to get near-realtime statistics. While having service VMs on a promiscuous portgroup may not be palatable to all tastes, I find it to be the better way to get live and continuous information.
Founder and CTO Jagan Jagannathan states “the only constant is constant change”
Do You PCoIP?
Xangati recognizes that VDI monitoring is important, and no one out there is paying it much attention. Market grab! Just check out how pumped Dwayne Lessner (fellow delegate and VCP-DT holder) at IT Blood Pressure is for their VDI integration, and future plans to provide user based performance profiling.
Seriously, though, it’s very hard to find good statistics on PCoIP traffic for a View deployment. VMware has done a great job at making this data available via WMI / Perfmon, and Xangati is taking advantage of it. During the presentation and live demo, we saw first hand the latency, bandwidth, and dropped packets of PCoIP sessions. It’s a great start to finding those annoying user experience issues that tend to plague a VDI deployment.
Additionally, Xangati offers a Visual Trouble Ticket for end users to trigger a recording of their desktop when problems occur. Essentially it’s a way users to raise their hand and say “this is slow” and have the tool capture the session data (including processes) for an admin to later troubleshoot. Unless you like being on call at 2AM, which in that case just ignore this feature.
I just threw this photo in here to make any chocolate lovers jealous. Nom nom nom!
While I don’t exactly find the GUI of the dashboard a delight for the eyeballs, it does get the job done and is extremely peppy in regards to load speed from link to link. I think they made a conscious effort to make usability a priority over pretty graphics, and I salute them for it. The dashboard is also customizable (you can add or remove widgets from the page), and future builds will include the ability to better “pre load” the dashboard for new users based on their role. Xangati takes the stoplight approach for most colors, where a red palate is “bad” and a blue/green palate is “good”, making problem finding relatively easy. It has a very “Tron” feel to it.
The day concluded with a game show entitled “Wait Wait … Don’t Blame Me!”
I’d like to see some improvements on the UI look and feel, and less manual labor to deploy the product in an environment. Those are my only two gripes! As for the Xangati monitor itself, it seems very slick and feature rich, and I’m a huge fan of the “recordings” that you can do to later play back an issue, which can also be shared out to other departments (great defense!). You may also want to read Rodney’s post at Musings of Rodos (fellow delegate) or Brandon Riley’s post at Virtual Insanity.
As for their presentation at Virtualization Field Day 2, this company really gets it! The chocolate and bacon swag aside, their presentation was extremely light on PowerPoint (only some funny cartoons to make a point) and included a whiteboard session from their Founder, live demonstrations, and a login for each delegate to try out the product for themselves. It doesn’t get much better than that. Great job, Xangati.