System and apparatus for automatic health monitoring and maintenance of virtual machines (18-Dec-2009)

System and apparatus for automatic health monitoring and maintenance of virtual machines

A virtual machine, perhaps in a cloud environment, is used for some long-running task such as running a server or application. In the situation where the machine or applications running upon it go into a "bad" state (e.g. they stop responding), the system administrator will need to fix this. Currently this is a process that requires manual intervention (at various levels). In a cloud computing environment the typical solution would be to re-provision the machine from scratch.

    Existing solutions are quite intensive and time-consuming. This is partially due to the fact that state must be re-established from a lower level e.g. re-install or re-configure the disk image, operating system, middleware, and applications. This can be intensive both from a labour point of view and on machine resources.

    A further drawback is that all runtime state of the machine and applications is lost. This must be re-established and depending on the application this can take a significant period of time. For example, a system rejoining a data-processing cluster may need to re-download the entire data set before commencing work. A freshly provisioned system can also take a significant period of time to "settle in" before its performance reaches optimal levels.

    Some automation can be used (particularly in the field of cloud computing) to script this re-establishment of service. However, this only removes the manual intervention part of the task and can still impact heavily on machine resources.

    The system monitors system health, and automatically creates "known good" snapshots and automatically reverts to these when acceptable health criteria are not met. The s...