If you are considering server virtualisation, then these are the 3 platforms you will no doubt be reviewing: VMware, HyperV and Citrix Xen server. But also bear in mind that by consolidating server processing on fewer physical hosts, resilience MUST be considered at the same time. All the eggs are in a smaller basket so if there is a component or full server failure then multiple server applications will be affected.
So what are the differences?
The common mistake people make is that VMware is virtualisation. True it has the largest market share – and it was the leading pioneer in taking servers to a virtual platform, however it is not the only way to virtualise servers and by no means the most suitable in many cases. VMware was developed before processors became virtual aware (or VT enabled – in Intel speak). Therefore the VMware hypervisor does ALL the work and cannot utilise the already virtual capabilities of the more modern processors. One major issue with VMware is system fault tolerance – it is not a proven solution with very little adoption. Full system fault tolerance cannot be achieved using VMware – and in all cases, anything but a basic failover solution requires the use of shared storage, which in itself is a single point of failure.
HyperV is actually Citrix Xen server architecture with a different badge and made more “Microsoft” interface friendly. But where it actually lacks behind Citrix Xen server is the related resilience that can be applied. HyperV can only achieve a basic level of cluster type failover. This a reactionary technology – where something has to fail in the first place for another physical cluster node to react and failover (with a period of downtime during the failover). This also relies on shared storage – which is a single point of failure, unless SAN to SAN replication is introduced. This is then becoming quite a complex and costly configuration.
On the other hand, Citrix Xen server is a free hypervisor – and system fault tolerance can then be applied using Marathon’s everRun MX solution. This means that if a component fails – or even a full server host fails – there is zero downtime. This is not a reactionary failover – it is true zero downtime. Also, no shared storage is required, hence total cost of ownership is typically less than a solution that cannot achieve such high levels of resilience using HyperV or VMware. There is also no operating system required at physical host level. Citrix Xen server creates the virtual machine and a windows operating system just needs to be loaded within the VM. Each bare metal machine uses Xen server as the underlying operating system, which is actually a Linux kernel, thus far more stable and less prone to attack than hosts with Windows O/S loaded at bare metal.
Both HyperV and Citrix Xen server utilise the virtualisation elements of VT enabled chip sets, hence are much more efficient and less tied down by the basic fundamentals of handling processing environment virtualisation.
Need more information? Please feel free to contact us to find out more.