Can we use Ceph or Gluster FS to avoid IO bottleneck to start many VMs?

0

AIM:
I want to start 50s of VM on s single server or on a set of PCs.

Scenerio:
I tried to boot 15 VM on a single server with enough computing power(processor) and RAM but disk is a bottleneck in that case. After searching a bit on google I got to know about Glusterfs and Ceph FS. I am using template images to start many VMs from those template images using overlay concept i.e single qcow baseimage and an overlay for each VM instance. I am using KVM as a hypervisor and Ubuntu as guest and host OS.

Problem:
Now I am using 10 PC with enough CPU and RAM to start 50 VMs. I can't boot 5-5 VMs on each machine manually. I am able to start all these VM using libvirt remote connections. I tried to mount the disk on all the machines using NFS. But as you can guess that will be too slow. So all 'n' all, I want to make it automatic and want to use same template file to boot all those VMs. So Is there anyway to use glusterfs or CEPH to use disk of all these 10 machines and satisfy IO need for these many number of VMs.

A-B

Posted 2015-05-26T09:49:53.900

Reputation: 111

Spread them across several drives to get more I/O. In particular get SSD (solid state disk) like the m500 or m550 and get 500mb/s read and disk IO bottleneck solved. – cybernard – 2015-05-27T01:13:21.783

@cybernard I want to use same template file as a base file to create VM instances by giving derived image file to each VM instance. So I can't spred across the disk image files, it should be central to each machine. – A-B – 2015-05-27T08:35:54.143

No answers