Xangati review

Many monitoring applications work across the entire universe of systems simply because of the protocols they use. After all, if you write an SNMP-based monitoring application, it's just as capable of monitoring (say) a Linux server as it is of probing a WAN router. Xangati takes a different approach, however, and sits within a specific vertical market: the virtual server infrastructure. Hence the Xangati product range sits on a VMware platform and hooks into that platform's on-board OS and virtual devices (eg.vSwitches), focusing its functionality and reporting on that specific aspect.

The product range is, therefore, no surprise: a range of “dashboard” products that shows, in real time and in on-demand reports, the specific behaviour of your virtual servers and the hosts they sit on. The two core products are the VI Dashboard (for monitoring a server-orientated virtual setup) and the VDI Dashboard (which gives more focus to virtual desktops in the virtual infrastructure), along with the StormTracker storm analysis tool (more about which later).

Although they're actually based on CentOS Linux 5.4, the Xangati applications ship as VMware virtual appliances. This is my favourite approach, as it eliminates any uncertainties of compatibility, configuration, enabled options and patch status of any underlying operating systems. Deployment is therefore a simple case of hitting Deploy OVF Template in the vSphere client app and waiting a few minutes for it to upload and become active. I used the VI Dashboard, as my lab is focused on servers rather than virtual desktops.

Once you've powered up the appliance, you need to log in and configure the basics. This is a simple matter using the CLI interface via a VMware console window; first you set up the IP configuration, and then you walk through a quick Q&A session about timezones and NTP servers before the device restarts itself and you're ready to go. As a result the device will be able to talk to the network, so everything from here on can be done via the web GUI.

Dashboard

The first thing to do when you log in at the web GUI is walk through the wizard that lets you identify your ESXi hosts and vCenter server(s). You then select which, if any, of your devices will deliver NetFlow information (if you have separate NetFlow probes you may not want to pull that information from the ESXi infrastructure itself) and then you have the option to automatically deploy the Xangati Flow Summariser (XFS) appliance to the infrastructure. The XFS, incidentally, is a small VM that hooks into the VM infrastructure's virtual switches and provides traffic flow information to the dashboard application for analysis.

The default view of your world is the Dashboard pane, which you can customise to show you the data items you're interested in. I counted over 50 different options in all, and you have the ability to group by nine different categories (eg. by user, by application or by data store). It's a quick job to customise your default dashboard so it gives you a nice broad summary of what you see, and you can also create custom dashboards for specific purposes (e.g. the storage team may be given one that focuses on the storage-centric items and doesn't worry to much about application focus).

From the dashboard screen you can click on most items to be taken to the Monitor page for that entity – be it a VM, a vCenter server, an ESXi host, etc. One of the nice features is the ability to initiate “Recordings” of traffic, so as well as the real-time monitoring and alerting you're given the selective ability to record traffic on particularly interesting items for later analysis. The applications, incidentally, work on the principle of “profiling” your systems – so they'll analyse usage for a specified duration to establish what “normal” behaviour looks like in order that alerting can be done in a sensible context instead of just using arbitrary thresholds (in system monitoring the word “threshold” is, of course, synonymous with “wild guess”).

Because the package sees across your virtual infrastructure it can warn you via its “Capacity Planning” section about capacity issues – so along with reactive alerting based in the current reality it's clever enough to use what it's seen to project requirements in both the short term and the long term. Finally, they've included a Visual Trouble Ticketing system which, at first glance, seems like a fairly dull “let's chuck a ticketing system in” concept but which is in fact a bit cunning: when a user raises a ticket the system automatically starts a traffic recording on the relevant flows – which is an intriguing compromise between recording nothing (and making life hard to diagnose) and recording everything (which is impossible, storage- and network-wise).

The documentation is also interesting. Although there are installation guides you can download in PDF form from the website, the company has produced a variety of walkthrough videos and published them on YouTube. I was initially sceptical but in practice I loved them and they were super-helpful; I was also much amused by the fact that the guy presenting one of the install guides says the phrase “go ahead” three times in every sentence.

The user guides are downloadable PDFs which you get via the Help menu of the application itself once it's up and running. They can be described in one word: dreadful. The English is annoyingly appalling, but more importantly while it tells you how to use a particular feature, it's sketchy on what that feature is actually for and why you'd use it.

The Xangati suite feels to me like some techies have had a blindingly good idea (a vertical application that probes deep into VMware's innards instead of something that just picks at SNMP objects) and put together something that lets you see that data. What it needs is a decent writer to sort out the documentation, and a GUI designer to work out a way to make the screens less busy (I've seen Xangati running in a real setup, and the average dashboard screen is way too full of numbers to be any use – it needs either a high-level overview with the ability to drill down or a set of opera glasses so you can read the tiny print on the wall-mounted monitoring screen).

The StormTracker app (which the company says lets you investigate “storms”, but which in my opinion really deals with less exciting sounding “performance problems”) is actually heading along the right lines. You get an overview (again based on the same VMware low-level probing and recording features of the other bits) that shows you where the problems are and you can drill into an unhealthy-looking area to not only see what's alerting but more importantly to look back through the recordings and see what's happened historically. So if one of your hosts has been maxing out it gives you a quick view into which VMs are running on that host, and lets you wind back through the recording to discover that a particular process on a particular VM has been eating resource at the appropriate times. Very neat, actually.

The Xangati family is really very promising. It shows you what you want, and it solves a key problem with VMware – namely that the latter has built-in performance monitoring and reporting but it's bloody awful to use. Xangati's alerting is useful and clear, its recording feature is cunning, it's sensible enough to use profiling instead of thresholds, and StormTracker demonstrates that they do really understand how people like me want my performance issues reported and presented.

Verdict

Xangati has made a solid start here. If it can sort out the documentation and make the GUI less cluttered in the next version, I think they'll have a properly attractive product.