How improve efficiency of server usage, reduce energy requirements and increase efficiency of storage resources in Datacentres.
Datacentres are one of the most power-hungry and energy-wasteful of all the IT components in a business. Typically formed from the integration of different applications and departmental systems, they often have as much as 3x more storage and 6-8x more servers than is actually needed.
As well as being damaging to the environment, this creates a huge challenge to businesses in terms of cost, power, agility and resilience. However, these problems can be eliminated by taking an architecture approach to Datacentre design.
The Problems
Datacentre servers usually only operate at about 15% efficiency. For every watt used by the server, another 2 go on air conditioning while one more watt is used up in lighting, power conversion and network components. These numbers might not sound like great cause for concern but multiply them by a thousand servers which is not even a large Datacentre by today’s standards and you'd be looking at a total power consumption of around 2.5 megawatts.
This level of inefficiency is so great that it is not just an environmental issue. Gartner claims by 2008 that 50% of today’s Datacentres will have insufficient power and cooling capacity to meet the demands of high-density equipment. By 2009 Gartner states that energy costs will emerge as the second-highest operating cost (after labour) in 70% of Datacentres worldwide.
The operating costs for an average Datacentre already run to around €10 million a year, of which €6.5 million is needed to cover cooling and UPS's, €1.75 million to power 15% efficient servers and €875,000 for 30%-efficient storage systems. Another €875,000 will be spent on powering a network with many appliances underutilised but drawing power and requiring space and cooling.
If current Datacentre energy consumption trends continue, Gartner has predicted that 50% of large organisations will face an annual power bill higher than their yearly server budget.
The Solution
It is possible, and indeed easy, to take a tactical approach to reducing Datacentre power consumption and carbon emissions: leave the facility essentially as it is, and trade up to more efficient servers, storage, racking layouts, cooling and power systems, and so on.
Such an approach would lead to small improvements, but planned equipment efficiency gains on their own will not be enough to curtail the overall growth in Datacentre power demands.
Furthermore, basing your carbon reduction approach on the power ratings of your devices alone can often hide the real problem, since a device which consumes twice as much energy might actually perform four times as much work, and thus be more energy efficient.
A much more effective solution to he problem is to take an architectural approach which looks at the totality of your Datacentre resources and tries to use them in the most efficient way by virtualising resources.
Virtualisation can help reduce network appliances and increase utilisation of storage and servers, essentially having a major impact on all equipment. It works at the server level by allowing multiple applications and operating systems to share servers that were previously only used for single applications, freeing up underused server capacity so that it is available for use with as many applications as needed.
This means the utilisation of individual servers can be raised to 70% or more, which could equate to a reduction of up to 75% of the servers in the Datacentre, freeing up space, power and cooling resources.
Within the network, virtualisation can help to reduce the ‘appliance load’ associated with each server; in other words, the number of firewalls, load balancers and other devices attached to it, which in a traditional appliance-based approach can represent an additional 700 Watts to one kilowatt of incremental power per server group.
This load is much greater (typically double) in redundant configurations, but in either case the load can be greatly reduced by using the concept of ‘virtual appliances’ on the network, simply building the capability of virtual firewalls, virtual load balancing, intrusion detection and other virtual appliance blades onto network switches.
In this way, 100 load balancers could be contained on a single blade instead of using 100 rack-mounted appliances.
Finally, virtualisation can be used across storage area network (SAN) fabrics to improve the utilisation of storage devices.
Previously, SANs were attached to individual application groups. A number of servers would have been attached to each individual SAN and the SAN would connect these specific servers to the disks in that SAN.
The challenge would be when one SAN was fully loaded and another had underused disks. The SANs could not simply be shared and were effectively ‘hard-wired’.
Now the same can be achieved with fewer SAN switches, equipped with virtual SAN capability and inter-virtual SAN routing, enabling any server to be allocated to any disk and allowing virtual SANs to be set up as software configuration exercises, in the same way as LANs have now progressed to virtual LANs.
This will lead to an increase in the utilisation of disks of around 70 to 80%. Fewer disks are needed and hence there is less of a requirement for space, power and cooling.
The fact that virtual SAN capability leads to a more flexible architecture means a single virtualised system can take the place of many, many more first-generation SAN switch devices, reducing switches, improving disk utilisation and cutting the total power required by up to a factor of three.
Furthermore, thanks to innovations such as Wide Area Application Services, virtualisation can be extended to include distant branches, consolidating underutilised branch office equipment into a super-efficient virtualised Data Centre.
The Benefits
Using architecture-based virtualisation, your business can achieve:
- an increase in server efficiency from around 15% to approximately 70%;
- a reduction of up to 85% in network appliance energy requirements; and
- up to around 70% utilisation of storage resources, from traditional utilisation rates of around 30% or less.
You an also remove the need for power-sapping redundant configurations, since through virtualisation it is possible to maintain business continuity while running two live storage centres, rather than having to have a backup centre always running idle.
Adopting an architecture approach involves some up-front investment in Datacentre technology however the savings generated are such that the payback time for the investment is usually only about 6 months.
Once you have the Datacentre virtualised and under control, you can then consolidate the branch office technology where it makes sense, continuing the power and cost reduction through consolidation and virtualisation.
Back to The Green Room Homepage
More articles:
Greening Data Centres Needs Realistic Risk Assessment
What is the CRC Scheme?