Questions[consolidation](server)

I have a customer that is in need of a platform upgrade and has a specific use case that makes me think twice about consolidation.

They basically have two services in the network that are currently not even isolated from each other.

1. They have a site in US and a site in Asia. The US site hosts a filesystem with thousands of files that is synced to Asia for "realtime" work on the files.

2. They host a VoIP server for a small call center.

So in turn, the CPU hungry process of syncing was starving the VoIP services and degrading call quality and even dropping calls.

We were able to somehow limit the sync process to at least not starve the FS services but an upgrade and isolation is of tasks is required.

Both FreeSWITCH (VoIP) and rsync can be very parallellized and will benefit of more cores but the use of virtualization for the box seems not very advisable because of timing issues and audio quality.

In short, whats your opinion:

More cores and kernel level separation of the tasks? Two separate boxes?

I've got a small virtualised environment which hosts a small internal website with database backend. I've currently separated my webservers and database servers on separate VMs running on the same hardware.

However, I'm wondering if a better strategy might be to consolidate the dev/prod database servers onto the same VM(s) but using separate instances? I'm using SQL mirroring for redundancy, so I'd theoretically go from 4 SQL VMs down to 2. Advantages would be fewer VMs to manage, less complex resource allocation and a guarantee that both dev and prod are on the same patch/service pack level.

What are the downsides to this approach? If I install a default MSSQL instance on a Windows server, can I just run the installer again to create a second (named) instance and use that for the other environment? Finally, would it be possible to put different instances behind separate IP addresses if I allocated both to the same VM?

In Exchange 2010, you can define Database Availability Groups and have your mailbox database(s) automatically replicated between multiple servers, and you can have mailbox server which are members of a DAG also host other server roles, such as Hub Transport and/or Client Access (which wasn't possible with Exchange 2007).

But can you define an array of Client Access Servers on the same two servers which are members of a DAG? The documentation I've found states quite clearly that you can't use NLB with DAGs due to incompatibilities with the cluster service... or, that you can try to set it up, but it would be an unsupported scenario.

What about using a hardware load balancer? Or a reverse proxy which can publish multiple internal servers on the same external URL (such as ISA)? If NLB is not used on the two servers and another balancing solution is implemented, can then full high availability of all main Exchange services (Mailbox, Hub, CAS) be achieved with only two servers?

The biggest benefit of virtualization is usually said to be improved server utilization.

But why do I need virtualization for that?

Say I got N physical servers that are lightly used. Why don't I just combine all the apps on those N servers into 1 physical server? This way I don't incur the performance penalty of virtualization.

What does virtualization buy me in this case?

I'm looking forward to getting your thoughts on consolidating/virtualizing 3 Windows 2008 Servers and 2 Linux Debian Servers into 1 (powerful) machine.

What is the most cost-effective Virtualization software available to accomplish this. VMWare looks awfully expensive!