Tags


Latest Posts


Latest Comments


Authors

Utilising Virtualisation For Disaster Recovery

stefan-haase.jpg

Posted by |

According to Gartner, around 15-20 per cent of IT workloads are already virtualised, and this figure could potentially treble to 50-60 per cent by 2012. Such positive adoption is easily explained when you consider that the resultant benefits read like an IT director’s wish list. A typical virtual machine environment brings significantly reduced hardware, energy and space costs – in terms of power and cooling alone, virtual servers are reckoned to be 700-800% more efficient than a traditional server. As well as the cost and environmental savings associated with consolidation, virtualisation can also help strengthen operational resilience and data security; expedite the provisioning and scaling of IT resources; enable the rapid deployment of applications; and streamline back-end administration.

A recent virtualisation project by international law firm Irwin Mitchell affords a perfect illustration of its potential. The firm was able to house 143 virtual machines on 11 physical servers, allowing the removal of 132 physical servers from its data centre; at the same time it saved an average 201 days of the provisioning cycle that would have been attributed to hardware lead times and 64 days of resource effort in racking and configuring the servers.

The London Borough of Hillingdon now runs 94 virtual servers on just three physical machines which it claims is a 97 per cent reduction in server hardware – saving it £20,000 per year by reducing power consumption from 34kWh to 1.1 kWh and an extra £50,000 per year by removing the need to expand its IT team. There are many more projects out there that prove the theoretical benefits do indeed translate into reality, providing plenty of encouragement to those still ‘thinking about it’ to get on and do it. But while the early majority ponder and plan, the early adopters are already focusing on the further leveraging of virtualisation technologies. They are starting to look beyond their original software development and server consolidation credentials and assess their suitability for one of the most critical disciplines within any modern day organisation – business continuity.

The starting point is to look at the problems associated with traditional business continuity and disaster recovery strategies. At enterprise, multi-national level, it is estimated that almost 80 per cent of the Disaster Recovery budget is given over to the protection of just 20 per cent of the total server network, namely their most business-critical servers. It’s an understandable risk management trade-off.

In an ideal world, organisations would want to protect all server workloads, with Disaster Recovery procedures geared to rapid recovery with a high level of data integrity – they would look to a separate data centre or centres and maintain mirrored environments, perfectly synched and ready to go. But the reality of those hot standby environments, and the high end server clustering and data replication that powers them, is that you need a one-to-one hardware relationship and software redundancy: you must have exactly the same server configuration at the discrete data centre as at the primary site with precisely the same operating system versions, licenses and patches installed. It’s a costly and a complex undertaking that is justifiable only in relation to ultra mission-critical servers; the rest of the IT estate will have to make do with more sustainable but lower – and therefore more risky – levels of protection. Think of it in insurance terms: you arrange a reasonable level of cover at a reasonable price and you accept the risk/cost compromise; it’s only when you want belt and braces cover for those specified high value items that the premiums shoot into ‘how much?’ territory.

Against that backdrop it’s easy to see what triggered the interest in virtualisation as a new Disaster Recovery dimension. Even by the crudest measure, replicating five production servers on a single server running multiple instances of a virtual operating system is going to reduce costs – not a directly correlating 80% reduction and not without a certain investment of time and skill in the execution, but a significant reduction all the same. Its this inherently efficient portability that has got people excited because here at last is a way to break the stranglehold of the one-to-one server relationship.

Portability is the essential by-product of the real ‘virtue of virtualisation’ – encapsulation. You can encapsulate an operating system, software, and its data into the equivalent of a single discrete file, or workload; you can then transmit this workload to a secondary location, create a bootable backup on a virtual recovery platform and get your users back online in a matter of hours, or less. Suddenly, organisations don’t have to be shackled by the heterogeneity of disparate servers, operating systems, applications and data sets – instead they can leverage the homogeneity implicit within portable workloads.

They can now move and replicate complete workloads as discrete, aggregated units, detaching them from their native hardware configurations and directing them between physical and virtual hosts as needs dictate – from physical to virtual, virtual to physical, physical to physical etc. One could argue that for a hard-pressed IT team the cost savings are very much secondary to the agility and flexibility afforded to them in dealing with unpredictable service disruptions. That’s the trouble with disaster recovery – most people have at least got beyond the ‘it’ll never happen to me’ ostrich act but find themselves stymied by an endless array of ‘what could happen’ scenarios and crafting a suitably risk-balanced response. Disaster Recovery can come out of the mundanely routine – the hardware or circuit failure, the software glitch, plain old human error – or the spectacularly devastating – the fire, the flood, the bomb, but either way you need to be adequately prepared for whether it is just one server that’s affected, a single site or the whole corporate wide area network.

Comments

 

Post a comment

Comment submitted! Comments needs approval before being displayed.