Virtual machines

Tuesday 13 May 2008

Chris Yearsley, Principal Consultant Xicon Limited


Chris Yearsley, coming from an academic and theoretical background but now working in industry, kindly offered to tailor his presentation to the needs of the audience, based upon our questions. He exposed some common misconceptions about shared storage and about virtualization.

The talk started with an overview of shared storage as shared storage and virtualization are interdependent technologies. Shared storage can be oversold yet underused once implemented. It should be enterprise wide and contain the organisation’smain applications and databases. It can also be used to reduce the time of backups as far as the application is concerned. One can also do snapshots of data, which could help with disaster recovery in certain circumstances.

When you virtualize everything, including the boot drive, operating systems and storage, is on the storage array. Snapshots are possible; the percentage of storage space taken by the snapshots differ depending upon the virtualization software used.

Virtualization makes the “hardware” generic so the organization may loose the advantages that it had recently paid for e.g. if you bought a particular server because of one hardware aspect then that reason will no longer be applicable. As a result of virtualization an organization should be able to alter the way it uses its systems, rather than just saving money. The rest of the talk only covered Intel virtualization.

What is virtualization?

It is invisible to the end user. You can run multiple operating systems and have different IP addresses etc. One or more computers are simulated; to the end user it appears that they are working on “their” server.

If the virtualization is set up appropriately then it can reduce outages and can go into automatic failover in two minutes (compared to a “real” server with a service level agreement of hours). This increased reliability can help to sell the concept.

When looking at disaster recovery, if the hardware of the main server is not identical to that of the recovery server then one can encounter problems (e.g. because the operating system is dependent upon the hardware) - but disaster recovery needs to be fast and have few risks associated with it. Compare that with disaster recovery in a virtualized situation where one can ensure that the virtual hardware of the two systems is identical - even down to the IP addresses used.

Even though virtualization uses software partitioning (very flexible) as opposed to hardware partitioning you can still get SLAs on a virtual server.

VMware, which is one virtualization tool, makes everything into one (rather large) file. This file can be copied. This means that you can move an application during production without the user noticing. For example, you can move it onto another physical host, repair or upgrade the original physical host then move everything back without disruption using VMotion.

If a host fails then VMotion will redistribute the applications across the spare physical capacity, taking account of priorities set against each application - so long as everything is on the same Storage Area Network. It is even possible to have a virtual infrastructure.


This gave us all an insight into the capabilities and power of virtualization and some of the benefits that could ensue.