Imaginary Computing - Virtualization

There was a time when the term virtualization perhaps conjured up images of bulky goggles worn for immersive "real world" simulation, described imaginative worlds, or somehow referred to the process of creating your Second Life avatar. For those not technologically minded you would be forgiven in thinking this somehow related to imaginary or pretend computers. However, virtualization has become a godsend for IT professionals, data centers, and cost accountants.

Virtual computing is becoming as ubiquitous as the very physical computers they sit on top of. It would appear that everything in IT can be virtualized today: servers, desktops, communications, applications, storage, databases and IT personnel (well, maybe not yet anyway!). In true IT fashion there is also an accompanying bevy of jargon such as hypervisor, VM and LUN.

While there are many forms of virtualization, the fundamental principle remains the same: abstract logical use from its physical use. This concept largely describes software independence from its hardware. It therefore infers that multiple software can use the same physical hardware, or conversely a piece of software can use different hardware.

Virtualization is not a new concept having originated back in the 60s on the mainframe. Back then it allowed for multiple operating systems to run simultaneously on a single mainframe. As previous real machines only ran one system, this gave rise to the term Virtual Machine (VM) for each multiple system running on the same hardware. Retro gamers that run Atari or C64 emulators on Windows/ Linux would also be familiar with the concept as they no longer need their beloved hardware to run these games. To monitor and manage all the operating systems requests of the hardware a software layer called the Hypervisor sits in between. The following diagram perhaps makes sense of the different elements and layers.



Virtualization has now been somewhat reincarnated on PC architecture and while it has many forms the most prevalent use of the term is towards server platforms. This has largely been driven by the increasing divide between hardware capacity and software requirements of it. When you consider that server resources (processors/memory) have an average utilization around 15% there is a lot of idle capacity. It's akin to driving your Ferrari (virtual for most of us) in peak hour traffic. In this example you could theoretically fit 6 servers on to the one machine with 10% headroom. This server compression is known as the consolidation ratio (e.g. 6:1).

The immediate benefit that springs to mind is capital cost reduction, especially if you are on the verge of a server refresh or new implementation. From a data centre perspective this footprint reduction has immense value to its operating density (power consumption/floor space) by reducing energy costs both from equipment consumption and its associated cooling. If you take the concept of logical separation further, then storage should be distinguished from processing. Enter storage virtualization which allows specialized disk hardware to alleviate application servers of the need for large local storage (including OS). Rather than provide a dedicated disk, storage hardware can be broken up into Logical Units referenced by a Number identifier (LUN). This not only improves utilization of disk hardware but can also reduce the rack size of application servers leading to higher densities and thereby greater data centre efficiency.

Less physical servers also imply less management and thereby staff headcount or efficiency should be factored into TCO and productivity measures. More than this, a Virtual Machine can be duplicated through a term called a "snapshot". Previously, what would take hours to install an operating system can now be done in a couple of minutes. While this demonstrates productivity improvement it allows the business to become more agile. Product evaluation, prototyping, sudden business activity or requirements can be handled with minimal effort and may no longer require lead times for hardware procurement.

If hardware is required it can be outsourced (if self hosting) to numerous hosting providers with virtualization capability in less time than it takes to order the hardware. In fact many hosting companies provide this scalability as part of their product offering in which you can leverage shared hardware should capacity exceed your infrastructure. This also works well for peak business activities such as marketing campaigns. Rather than carry the cost of equipment to cater for business spikes, one can simply "rent" additional capacity for the event.

Virtualization also introduces higher availability and recovery of digital assets and systems. It is easier to create duplicate servers for load balancing or fail over. Applications can be isolated in separate VMs, so should one fail it does not have to impact the use of other systems on the same physical server. The ability to snapshot and move virtual machines makes light of equipment maintenance and disaster recovery preparation.

Geographic redundancy also becomes accessible to mainstream business. Previously, this would require a mirrored environment at the recovery site. Now, for a fraction of the cost, a hosted provider can offer its shared infrastructure as a point of fallback. Data will still need to be synchronized or restored and sufficient bandwidth is required, however this facility provides near real time recovery.

For most of the reasons above this technology became popular for mainframes in the 60s. Funny how sometimes we come back full circle in IT. While the benefits are fairly clear, no technology is infallible and there are a few things to keep in the back of your mind when you virtualize.

The hypervisor may have limitations. Some versions only allow a maximum number of processors/cores per Virtual Machine. While most applications should never exceed the restrictions, some complex systems such as database servers can. Do your research with the software vendor or ask the question of your hosted provider.

Don't rush off and virtualize everything! Server consolidation is based on the premise of idle capacity. Intensive applications reduce this potential and some may be better off on dedicated hardware. Furthermore, average utilization is not the same as peak utilization. This is an important factor when consolidating as to strike the right application balance so that peak utilization does not breach capacity or degrade performance. Be wary of virtualizing database servers. Not only can these be resource hungry, they are also disk intensive. Storage virtualization adds a degree of latency and may impact on performance requirements.

Virtual redundancy does not entail physical redundancy. Hardware and components should still have a measure of redundancy to fail over on mechanical failure. For instance, if you have six servers and they can be consolidated to one, you will probably want a second as a redundant measure. In addition, duplicate virtual machines for high availability purposes should straddle multiple physical servers where possible.

Software is fallible and ultimately the hypervisor is still software. It can be subject to corruption from physical failure and human programming error. Of concern will be security in the future as malicious code seeks to exploit any vulnerability here. A successful exploit would jeopardize multiple servers rather than one at a time. While the risk is relatively low, a security strategy should be implemented. Physical separation may form part of this, but more security software is emerging to sit between the hypervisor and its Virtual Machines.

If you decide to outsource hosting there are many providers that offer virtualized services. There are a few things to keep in mind here. The first may seem obvious, but virtualization should not cost more than dedicated services! Furthermore, a hosted provider achieves economies of scale and the TCO should be lower than providing the same services yourself. Many will charge for both hardware and Virtual Machines. While there is licensing costs to recoup in the VM charge, be wary of double dipping. There certainly should not be a separate hardware charge if you are utilizing shared infrastructure.

Virtualized services can be provided on your own dedicated equipment or alternatively leverage shared hosted infrastructure. The latter definitely offers cost savings but it is important to ask the vendor of its contention ratios here, that is, how many other services share the physical hardware. This is a similar ratio for bandwidth contention, the lower the better (e.g. 2:1), while the converse suits the vendor (e.g. 40:1). The concern here is over gearing so that performance degradation surfaces at times. A good service level agreement is important but getting it right from the start is always best.

Virtualization is now established technology with a clear value proposition. In a world of climate change and a focus on carbon emissions, virtualization also has a beneficial role to play. Server consolidation can lead to less manufacture which has flow on benefits to carbon output and land fill. Virtualization can certainly be a positive move in good corporate citizenship in conjunction to the business value it derives. However, like most things in IT, do your research, scope out your requirements and plan effectively, otherwise the imagined benefits will virtually disappear in front of you!



About the author
Sheldon has been in the IT industry for over 15 years and held senior leadership and strategy roles. He's worked within a number of notable corporate giants and more recently in sports/entertainment and green tech industries. His forte and contributions here revolve around the web and cloud computing along with digital experience.