Data centres: Extracting more from less

Cloud adoption should be a medium term strategy spread over three-four years, spanning across hardware and application refreshes, notes A S Rajgopal, MD, NxtGen DataCenter & Cloud Services Limited

Excerpts:

CIOL: What are the challenges that CIOs face in data centres with cloud mode of operating?

Rajgopal: In enterprises, core applications run on a stable platform. So, CIOs will have to take a call on whether or not there is a requirement to port them on cloud/virtualised infrastructure.

They will have to create spare capacity on existing hardware to cater to growth/performance needs. They also have to check if the cost of cloud adoption delivers the economic value it promises and whether their new hardware is cloud optimised. They also face other infrastructure related challenges.

CIOL: What needs to be done to address these challenges?

Rajgopal: During the C-Change event it was commonly agreed that cloud adoption cannot be an event but it is a journey. Cloud is being adopted to ensure that we are able to get 20- 30 per cent additional performance from hardware by smartly utilising idle times and performance head-rooms. The cloud adoption should be a medium term strategy spread over three-four years, spanning across hardware and application refreshes. All new deployments should definitely be enabling cloud adoption strategy.

CIOs should avoid virtualising such application servers which are already running at 60-70 per cent average utilisation, attempts to get more out of such deployments will not yield any gains. Similarly, there may be some hardware which is ageing but under-utilised. Virtualising them may not yield any measurable results. The big bottleneck in virtualising is disk performance - availability of disk space does not mean that it would be useful for virtualised environment. Most cloud deployments are based around centralised storage based on Fiber-channel/iSCSI/High performance NAS etc.

In short, all the available hardware need not be virtualised into a cloud. Careful analysis and detailed adoption strategy is essential prior to embarking on the cloud journey.

There has been a significant change in the way IT hardware is designed in last couple of years. Earlier, servers used 60-70 per cent of their total power even during their idle time, today they use 20-30 per cent. Being energy inefficient the older hardware may be costing far more than going through the effort of virtualising it to be 20-30 per cent more performance headroom.

On the other hand, processing is not scaling up, but scaling horizontally. We now have more cores (processors) than a faster processors. The only way to best utilise this modular computing is by adopting virtualisation. Economic value of virtualising more recent hardware is greater than the earlier hardware.

We have seen that due to power utilisation inefficiencies, capacities within the data centres are becoming a bottleneck. CIOs should look at improving their data centre energy efficiency by adopting greener technologies to not only save operating costs, but also increase capacity.

Today it is possible to revamp data centre and the cost savings payback in couple of years, unlike earlier days where data centre investments were planned for seven years or more.

CIOL: What should enterprises do to tap the unused resources in data centres?

Rajgopal: A data centre has IT and non-IT assets. It is important to understand areas where more can be extracted. A mature DCIM system can gather information from both IT and non-IT assets and provide the required information to assess areas that can deliver more.

Power being the biggest cost for a data centre, we should start with measures to save power, thereby creating more capacity. Gather information on spare computing and storage and evaluate the cost of virtualisation against the additional performance gain, and go for strategic deployment of private cloud. Non-core applications, which do not require guaranteed performance needs can be ported on available public cloud resources or look for hosted private cloud resources, which can guarantee performance.

CIOL: Today every other player is a cloud service provider. What will be your differentiating factors?

Rajgopal: At NxtGen, we are focused on Enterprise Cloud Services (ECS) - a typical cloud services provider makes money by over provisioning the available resources and hope that all the resources are not demanded for at any given instance. We are not in the public cloud space, we offer hosted private clouds, which offer guaranteed performance. For us cloud is not virtual, it is physical. We offer virtualised, but dedicated hardware resources and customer gets to decide the number of virtual resources based on his own performance requirements. For us, economic value of the cloud is core deliverable, we have adopted OpenStack which is supported by industry leaders.