0

Ok, I understand generally about the clouds... I've been using virtualization of my servers for the last 3 years or so. Originally with vmware server, now I prefer xenserver...

Although my machines are no longer permanently tied to the hardware, I still have to watch my provisioning closely, making sure I properly match the loads, don't over provision, etc.

I still have a problem with migration because my servers aren't the same, so I can't move from my older intel xeon's (PE1850) to newer generation (PE1950) when using xenserver (yes, vmware can do this)

But now there's cloud.com, openstack, eucalyptus, etc. so I can run my own clouds...

Will these tools allow me to take my physical hardware and make them all contributors to the cloud, install xenserver to the cloud, and then OSes to the xenserver? When I need more processing/space/mem I just buy sufficient servers and add them to the cloud? This would make provisioning a lot easier, and eliminate the problem with not having matching servers...

If not, what would these tools give me?

3 Answers 3

5

There is no singular "cloud". It's a marketing term that has little to do with the underlying technologies; more and understanding that there is an abstraction between the "consumer" of the service and the people taking care of the infrastructure (to varying degrees depending on the service you purchase).

You can't "contribute to the cloud" because it doesn't exist in the first place.

These tools provide technologies which might be useful to you; if you were in marketing you would call the services they provide "cloud" related technologies.

2

I think you're curious about a 'private cloud', which in fact is very close to what you already have. The main difference is about having a single 'control panel' where you handle mostly 'high level' operations. These are some typical features of a cloud system:

  • a simple way to 'add hardware' (just installs an hypervisor and control daemons, and registers with the control panel)
  • management of VM templates
  • storage management (volume creation, deletion and assigning to VM instances)
  • VM lifetime handling (instantiation from a template, start, stop, statistics, destroying)
  • automatic migration between compute nodes according to load.

As you can see, you can manually do all this; the 'magic' is to have all integrated, and automating as much as possible. So you don't have to care about exactly which disks hold each volume, or which compute nodes run which VM instances.

So, is it for you? it depends, if you have many different VMs, you wouldn't benefit much about a templating system. Also, if some of your VMs do lots of data sharing, you might loose some performance if the load balancing algorithms decide to migrate those to opposite ends of your datacenter.

Where this setup shines when you have a large quantity of VMs doing similar work, so could be instantiated from a limited set of templates. In that case, it easily let you go from handling a few tens of VMs to several hundreds with the same team.

1

The other thing is, when you're the admin for the backend as well, it doesn't matter if it's a "cloud" or not. You've still got to admin the machines, do capacity management, and all the other normal tasks for any other server you manage.

2
  • right, of course; the advantage is that you can split in two teams (or two hats for a small team): one is the 'infrastructure' guys that handle the hardware and keep it running a single application: the hypervisor. the other is the 'application' guys that get the VMs to do whatever they want without having to bother about hardware details.
    – Javier
    May 7, 2011 at 1:20
  • You can also do that in a purely physical environment as well. At larger shops, I've usually seen it done that way.
    – mfinni
    May 7, 2011 at 14:03

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .