The 5th Usenix Symposium on Networked Systems Design and Implementation (NSDI’08) starts today in San Francisco, CA. The NSDI symposium focuses on the design principles of large-scale networks and distributed systems.
XenMaster Ian Pratt (Senior Lecturer, University of Cambridge Computer Laboratory/XenSource (Citrix)) will be presenting the Keynote.
Apart from this Keynote a very interesting paper about Virtualization and High Availability will be presented by the Department of Computer Science at the University of British Columbia.
In a paper titled Remus: High Availability via Asynchronous Virtual Machine Replication Brendan Cully, Geoffrey Lefebvre, Dutch Meyer, Mike Feeley, Norm Hutchinson and Andrew Warfield describe how with minor modifications to Xen 3.1.2
they achieve fast (frequencies as high as forty times per second) replication of Xen virtual machines.
Their tool Remus allows systems to transparently move to another physical machine in the event of a failure on the primary machine , with only seconds of downtime, while preserving the original host state such as active network connections , memory and disk state.
Today most Virtualization vendors claim they achieve High Availability while actually they only implement an easier way to recover from failure or a method to migrate away from a machine that is showing early problems. Most of the current Virtualization based solutions still require you to restart a new virtual machine, or implement High Availability on application level.
Mind , this is a research paper don’t expect these kind of features to show up at your local IT shop within the next couple of weeks.