Hello all,

We have a KVM installation on Cent OS 6.1 on a server with 48 gig of RAM. Everything is running decently and performance too seems fine.
But the issue is that there is no decent Management client for windows or web. Also, the virt-manager does not have enough information about the VM. I would like to have the information of VM in summary just like vSphere shows. Also would like to have the summary information for the whole server. Can anyone please help me out with a decent management console for KVM?

If you want a proper enterprise-level management solution, you need RHEV. If you prefer the upstream bleeding endge project - look at ovirt.org.

As for everything else, virt-manager runs quite well on windows through Xming and putty's ssh X redirection, and to view server stats there's a huge load of packages, from top/htop to webmin

We currently use virt-manager with Xming and ssh X forwarding. That works well for creating or deleting the VM. What we need is a proper management client that would give statistics about the server and the virtual machines running on the server. We usually over shoot some RAM allocation. I would like to have a statistics on that front. Like how much RAM has been allotted to each VM and how much is each VM utilizing. This is important stat as we could reduce the amount of resources allotted if needed. This allows better utilization of resources. We are open to using solutions that are paid to use. And if there are webmin modules that could help gather enough information regarding hardware resources of the server, I am open to trying out different things.

I've been using collectl for monitoring VMs inside the hpcloud infrastructure and it does a very nice job. For example, I can tell it to simply monitor all the processes with the name 'kvm' or 'libvirtd' and it will collect data on all the VMs. As an aside, the reason for looking for processes with the name of 'libvirtd' is that is the name the new vm starts with and a few seconds later it changes to kvm and collectl might miss it because once it decides it's not interested in process it never looks at it again - that's why filtered process monitoring is so fast!

Anyhow with this data being collected every 5 seconds (yes, you read that right) it's possible to see individual loads of all the VMs as well as how much I/O they're doing. Network traffic per vm is a little trickier because it's not directly contained in the VM process data. However, the string used to start the vm, which collectl also captures, contains the name of the individual vnet's MAC address. If you're also collecting network data you can then map individual vnets to the MAC addressing the VMs are using to accessing them.

-mark