Normally we have a one to one relationship between a server operating system and the server hardware it runs upon. That hardware will run only that
single instance of that operating system and the operating system is bound to that manufacturers specific hardware components (through the
drivers it runs).
As the software running on our servers gets ever more complex, there is a constant desire to have one server = one application. This makes
management easier and reduces the issue of applications clashing when they are hosted on the same server. We are even starting to see this
level of complexity with some workstation class applications. The downside of this is we typically end up with server hardware
that is capable of running many applications only running a handful or even just one. The reality is that many servers are
for the most part often sitting close to idle.
At the same time as this sprawl of hardware is increasing, there are constant pressures to purchase less hardware, which means
less equipment to manage, power, cool and replace.
Clearly the two scenarios are at complete odds with each other. Fortunately there is a solution - virtualisation. Better still,
this is a technology that has been around for a number of years and is now starting to make headway into the small/medium business
Instead of loading the operating system directly onto the hardware, a different operating system is loaded - the Virtual Machine (VM)
Hypervisor. The hypervisor provides a mapping layer between the hardware manufacturer's components and a standard set of hardware
components that it emulates (in software). The server operating system is then loaded on top of this hypervisor. However, the
key element of this hypervisor is that it is able to provide this emulated hardware not only to one installation of the operating
system, but to many. These instances of the operating systems are called Virtual Machines or VM's.
This immediately brings us two significant benefits.
Better Resource Utilisation
Now we able to what would have required multiple physical server machines onto perhaps just a single server. Each of these machines
operates in isolation, is completely unaware that it is sharing a physical machine with other servers and from the network, is completely
indistinguishable from a "normal" server. Each of these operating systems is contained in one single large file, meaning they
can be moved from one physical host server to another with relative ease.
Of course, some extra care is now required to ensure these VM's do not consume all the available resources (CPU, Memory, Disk) on the
The virtual machines are hardware agnostic. The hardware they see (disk controllers, network cards etc) is the hardware provided by the hypervisor and that
is the same hardware no matter what make/model of physical computer they are running on. With these VM's being contained in simple
files, we can quite literally copy a VM running on a Dell server to an HP server to an IBM server and they will run with no reconfiguration or
Rapid Server Recovery
Suddenly we have a whole new level of business continuity at our disposal. At the most basic level, we can backup these VM's and in the
event of a loss of hardware, they can be copied onto replacement hardware and restarted. No reinstallation of the operating system
or lengthy and complex recovery from backup. Moving hardware generations with the same OEM can be as complex as changing OEM. With
virtualisation, this is no longer an issue.
Server Virtualisation in the SME/SBE environment has some different drivers compared to the enterprise organisations where it originated.
The key driver for Server Virtualisation in the enterprise class of company was to reduce the number of physical servers. This in turn
reduces the space, power, cooling and maintenance costs. If you are an organisation with 30,000 servers and you can reduce this to 5,000
servers the savings are significant.
In the SME/SMB environment, while there are cost savings moving from say three servers to one, they are not so significant. There is
also the issue of now having all the servers on a single physical machine. Loss of one physical server really means the loss of three
However, even for an organisation with just a single server, making that server into a VM and running the free version of VMware (ESX/i)
gives us the benefit of the business continuity described above. We would now need to have a compelling reason not to use a hypervisor
even for a single server deployment.
Desktop virtualisation is built upon the same requirements as server virtualisation and upon the same solutions.
Rather than supply traditional desktops, these can be replaced with thin-terminals and the users remote desktop to their actual computer. These
computers are then VM's running on the servers. Thin-terminals typically have no moving parts and are very low powered, low cost appliances. They
are particularly suited to the factory/production environment.
Again, the enterprise class companies have their reasons for this sort of solution, but there can be benefits available to the SME/SMB organisations
as well. This is particularly useful for making legacy applications available without having to keep legacy hardware running. It can also be beneficial
where remote users need to access an application in the office and solutions such as Citrix are prohibitively expensive and overkill.
It is now possible to run a Storage Area Network (SAN) in a virtual environment, so the storage itself becomes virtualised. This can help reduce
the costs of implementing a SAN environment for VMware.
Exmos have recently implemented several instances of HP LeftHand SAN/IQ VSA which is a SAN that runs in virtual environment. For these installs the SAN/iQ VSA was hosted on VMware’s ESX/i (free) virtual platform.
In order to provide security between VM's it is now possible to run a virtualised Stonegate firewall in the VM environment.