[one-users] OpenNebula with Clusters + KVM

Keith Hudgins keith at cloudscaling.com
Thu May 27 09:16:38 PDT 2010


Here's some answers:

1 - Not exactly as you're thinking, but yes. You'll have all 10GB RAM
and all 5 processors available, but each machine is a discrete
resource.

2 - Good question, one I can't answer immediately.

3 - It's very possible. You'll need to run a shared storage config
with the ONEHOST directory shared via nfs or some other network mount
system, but it's not hard to do. Just make sure there's no
network/firewall barrier between your vm hosts on kvm's migration
port.

4 - There's a web gui available in the ONE ecosystem.
http://opennebula.org/software:ecosystem:management_console
You can assign multiple users and priveleges, but it's not very
fine-grained. Check the docs for more details.

5 - There's no baked-in fault tolerance, you'll have to do that
yourself. Running a pair of HAproxy vms with heartbeat enabled on
separate vm hosts load balancing between several app servers
underneath them, also on separate vm hosts is a good way to manage
that with very little extra work.

6 - Yes, but again, you'll have to roll your own. It's not
particularly difficult - you'll need to pull metrics from somewhere,
and write the triggers, but using the API to launch new vms is pretty
simple.

7 - Yup. It's just kvm (in your case), which runs must any intel-based guest OS.

Understand that ONE is a virtual infrastructure manager - it rides
herd on your vm hosts and gives you an api and a scheduler to allocate
new vms. That's pretty much all it does, so extra "cloud"
pie-in-the-sky services like autoscaling, etc are up to you to
implement. That being said, ONE is the most hackable VIM in existence,
and the API is very full featured, so with a little work, you can do
all you're looking for and more.



On Thu, May 27, 2010 at 5:07 AM, KING LABS <kinglabs.in at gmail.com> wrote:
>
> Hi ALL,
> I am new to cloud and opennebula. I have been exploring for best possible
> ways of implementing private cloud for my company.
> I found opennebula interseting.
> As I am new to OpenNebula I would like to know if the following use case is
> possible
> with OpenNebula and also request to suggest me the best way of meeting the
> following requirements.
> 1) Clustering:  Would Like to build cluster as single aggregated resource
> pool with kvm as choice of hypervisor.
> example: If I am having 5 servers with 2GB RAM & 2Ghz processor .... after
> clustering it should be consolidated and work as single
> system with 10GB ram and 10Ghz processing power for my KVM virtualization
> environment, where I can
>
> 2) Availability zones: Is it possible to isolate parts of the cloud for
> speci?c pur-poses or customers.
> 3) Live migration: Moving a virtual machine while it is still running.
> Migration of virtual machines is always possible when it is shutdown, but
> there would
> be a high level of ?exibility if this could be done while running
> 4) User management: Can users be managed and assign privileges to them. Has
> the platform a graphical user interface to interact with the cloud?
>
> 5)  Fault tolerance / fail-overIs there a feature to handle failing
> hardware? Running two instances of the same virtual machine (VM) on di?erent
> physical servers, when one physical server fails the VM on the other server
> will take over. The same can be done for physical machines by making them
> redundant.
> 6)  ScalingCould the cloud infrastructure adapt to requests. Is it possible
> to expand or shrink the resources automatically when needed or can this
> easily be done by hand (Automatically shutdown or turn on virtual/physical
> machines when needed).
> 7) Multiple Guest OS : Support for windows + linux mainly
> Thanks in advance
> Regards,
> KINGLABS
> _______________________________________________
> Users mailing list
> Users at lists.opennebula.org
> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
>
>



More information about the Users mailing list