 "Adventity Expands Operations With Virtualisation")
A full service KPO/ BPO firm catering to the BFSI and the airlines and travel industry, Adventity offers comprehensive knowledge services, transaction services and customer contact support solutions to clients across the globe. As an end-to-end service provider, Adventity combines research, analytics, processes and operations to benefit clients through seamless client integration, execution and scale.
Expansion in Operations Drives Need for Virtualisation
“We wanted to free up data centre space for additional projects, and knew that replacing physical machines with virtual machines would help us to accomplish that goal,” says Rajendra Sawant, CIO, Adventity. In the earlier scenario, CPU utilisation was low, and the servers required high electrical power and cooling. The IT group evaluated server consolidation strategies to increase the efficiency and reliability of the company’s IT infrastructure, while reducing cost and space requirements. “We looked at some workload management possibilities, but vendors required different servers. Finally, we chose VMware virtualisation because it was the only solution that met our needs,” says Sawant.
The Consolidation Estimate (CE) was the first step towards a Virtual Infrastructure deployment. The assessment was designed to provide Adventity with a snapshot of their current environment’s virtualisation potential by providing an accurate assessment of X86-based Windows and Linux servers.
The program included:
o Gathering an accurate inventory of hardware and software from current servers
o Gathering performance statistics
o Identifying server candidates for virtualisation
o Estimating realistic consolidation ratios
CE Prepares the Ground for Virtualisation Deployment
The outcome of a CE was a clear path to getting started with a successful Virtual Infrastructure Deployment. Building the CE took around three to four weeks. VMware Capacity Planner was used as a business analysis, planning and decision support tool to direct the key phases within a variety of infrastructure assessment projects. The highlights of the assessment included calculating the current workload capacity of the IT infrastructure through comprehensive discovery and inventory mapping of IT assets.
The system workload and capacity utilisation was measured across various elements of the IT infrastructure, including by function, location and environment. Once the capacity utilisation study was conducted, planning was done on how to make the best optimal use of the current capacity. Alternate ways for better capacity utilisation were also identified.
After the completion of the macro requirements, it was time to go into the specifics of identifying resources and establishing a plan for virtualisation, hardware purchase or resource redeployment. “We zeroed in on the optimal solution by evaluating various alternatives through scenario modeling and ‘what-if’ analysis and determining the best alternative that met the pre-defined criteria,” says Sawant.
In this exercise, resource utilisation was also monitored through anomaly detection and alerts based on benchmarked thresholds. This led to recommendation generation for ongoing capacity optimisation.
Challenges during Implementation
The foremost challenge was to select the servers for virtualisation. The planning and designing of the virtualisation deployment was a time-consuming process.
When asked about whether integration between the virtualised environment and the physical systems was an issue, Sawant says, “While designing the virtualisation exercise, the integration points were adequately planned. The hardware was also procured accordingly.” Thus, integration was not an issue.
Result of the CE: 36 Servers Virtualised on Three VMWare ESX 3.5 Servers
Through the Consolidation Estimate, the IT group was able to determine that the workloads of 36 servers could be moved to an equal number of virtual machines residing on only three physical servers running VMWare ESX 3.5. The virtual machines ran on three 4-CPU Dell PowerEdge servers with the fourth server as a back-up machine. The virtual images were stored on separate iSCSI LUNs (Logical Unit Numbers) created on EMC Celerra NS20 SAN, which provided the additional benefit of virtual ‘thin’ provisioning.
The benefits of this comprehensive virtual infrastructure solution include among others server consolidation. The IT group has been able to achieve at least a 12:1 server consolidation ratio, with 12-18 virtual machines residing per physical server. “We did a cost analysis standardising on 4-CPU servers. It was found that the number of servers could be stretched up to 20, but for production, 12-18 was the sweet spot,” says Sawant.
From a cost perspective, a 5:1 consolidation ratio is the break-even point in terms of hardware costs, thus, the company saves by not having to buy hardware. The other cost savings include air conditioning, electrical power consumption, rack space, server AMC and network ports. “If we hadn’t adopted virtualisation, we would have run out of ports. This year it would have cost us Rs 12,00,000/- to upgrade and add enough network ports to support our server infrastructure. We spread the cost over two to three years, so that means we saved Rs 4,00,000/- this year to defer the upgrade,” says Sawant.
There was significant cost savings achieved by doing away with the Asset Management Cost (AMC) of servers (IBM x345 & x346 series), which were out of warranty, and therefore chosen for virtualisation. This saved cost almost to the tune of Rs 7,20,000/- that would otherwise have been spent for the asset management of 36 servers. Electrical power usage went down drastically from 15 KVA to 3 KVA and data centre rack space was reduced at the ratio of 3:1.
CPU utilisation rate has increased from 15-20 percent to 50-60 percent per server. The other major benefit was the reduction in the server deployment time. Instead of taking four to six hours to set up a new server, it takes 15-20 minutes in a virtualised set-up. In terms of server procurement, ordering a new server could take weeks. With virtualisation, a new server can be deployed right away.
Virtual ‘Thin’ Provisioning is another advantage. This feature allows storage to appear bigger than it really is by only allocating storage when writes are actually performed. “Thus, the storage utilisation significantly improved and 3 TB space was enough to manage 36 servers, earlier, it would have taken anywhere around 8 TB with fixed storage allocation,” says Sawant.
The performance of the virtualised environment is being monitored since the last two months and we haven’t received any complaint as yet from business users. There are 18 more servers left to be inducted into the virtualised environment and the entire exercise is expected to be over by February end.