« A brief overview of load balancing techniques | Main | A Complete Cluster Stack for Linux »

13 December 2007


Feed You can follow this conversation by subscribing to the comment feed for this post.

Matt R.

Hi Alan,

many thanks for nice article pointing out the troubles and
needs to manage the virtualization hype :)

From what i read and understood the answer would be a pluggable
data-center management platform which is aware of monitoring,
HA and virtualization and additionally provides lots of transparency
and automatism to guide system-administrators and avoid human errors.
If your answer is "yes" than i would like to recommend to take a look
at the openQRM project -> http://www.openqrm.org

Many of your listed requirements have been already implemented in openQRM e.g.

1) support for large HA-cluster
openQRM provides an integrated resource-planning and provision system
which enables the system-administrators to create "clusters on the fly"
HA is available in 3 layers :
a) automatic hardware fail-over
b) application fail-over triggered by external or internal events
c) HA for the openQRM-server itself via one or more hot-standby's

2) integrate with many virtualization layers
openQRM provides a generic "partition engine" which conforms
different virtualization technologies within a single-management console.
Currently openQRM supports VMWare, Xen, QEMU (+KVM) and Linux-VServer
(but is not limited to that because of its open pluggable architecture)
In openQRM virtual machines are deployed and manged in the same way
as physical machines. Migrating from physical to virtual is easy
and completely transparent. That means a system administrator may decide
to run a server today on a physical system, tomorrow on a Xen-partition,
the day after tomorrow within a VMware-partition and the next day migrate it
back to a physical system without any changes on the server (-image) itself
nor with hasseling with the diffrent configurtion and disk-image files
from the different virtualization technologies.

3) Support monitoring and controlling services inside the virtual machine
Again, in openQRM virtual machines are managed and monitored in the same
way as physical systems.

4) Integrate with provisioning systems
One of openQRM base features is its provsion system. It provides a full
overview about all available resources in a data-center and how they are used.
New systems are being detected automatically and appear in the "available/idle" pool.
New virtual systems from diffrent types (VMWare, Xen, QEMU, Linux-VServer) can be
created (and removed) easily through a unified user-interface.
Also openQRM integrates with storage-infra-structure (e.g. NetApp, LVM2) to e.g.
automate creating/cloning servers (server-images) via snapshotting for rapid deployment
and/or backup/restore.

many thanks again + enjoy,

Matt R.

Raphael Bauduin

Just wanted to mention Matt's participation in the Profoss event on virtualisation, where there will also be a presentation titled "HA clustering made simple with OpenVZ" and one on management tools that could add some elements to the discussion.

All info at http://www.profoss.eu/events/january-2008-virtualisation/?tab=activities


The comments to this entry are closed.

Become a Fan

AddThis Social Bookmark Button
Blog powered by Typepad