xVM Server and xVM Ops Center 2.0
By nickstephen on Sep 13, 2008
What a ride! These past few weeks we've been working round the clock with our world-wide team to integrate and test everything, from the network discovery and provisioning of a machine that's just been unpacked and racked in some lab the other side of the world, through to the live migration of a virtualized machine running Windows, Solaris or Linux from one physical xVM Server to another.
As the architect of a lot of the backend infrastructure for all of this, it's
really rewarding (and relieving!) when the plan comes together, and when the result not only works, but surpasses expectations!
We'll be demoing the xVM portfolio next week at VMWorld, I'm flying out tomorrow to help on the demo stand.
As I mentioned above, my responsibility has been the backend infrastructure of xVM Server and xVM Ops Center, with the APIs that the built-in web-server uses to get its job done, and, behind that, the management data model that is distributed across a "cloud" of management daemons.
Each xVM Server has its own data model piece. So does the xVM Ops Center enterprise controller. If you deploy an xVM Ops Center proxy to offload some of the enterprise controller's work then you're deploying another piece of model too (you might want to do this to manage a lab on a different subnet on the other side of a firewall or simply for horizontal scaling for massive data centers), and you can even push down a low-footprint piece of model via a Java-ME based management daemon onto your Solaris or Linux operating system to delegate the system's monitoring and patch management directly onto the system itself.
To build this management cloud, the xVM products heavily leverage the Common Agent Container, which is an open source community project providing a JMX Management Services Container, enabling the easy deployment and administration of Java management modules in a secured environment integrated with the platform.
Each of the pieces of the data model is hooked together into a single model, with duplicate pieces of the model aggregated together so that you don't see the same system more than once. The connections between each piece of the model are done using the same technology you may be using to read this blog article - ATOM web feeds. This allows the distributed model to easily cope with corporate firewalls - you can set up direct https connections or go via your corporate web proxy, and also offers cool auditing and other integration capabilities since, with the right credentials, you can subscribe any network connection in the distributed model via any ATOM reader and monitor everything flowing through the system in real time from your desktop / phone-top!
At the same time it's great how the project I designed prior to this one - the Common Agent Container - continues to grow, going from being the mangement framework for Solaris Cluster through to the Java Enterprise System Monitoring Framework, into Solaris 10 and Nevada and Open Solaris and now as an important building block of our new and really impressive xVM management portfolio.
But it's not very sexy showing you data models, how about a screenshot from the upcoming xVM Ops Center 2.0 - click to get full size!
See you at VMWorld in Las Vegas?