Established in 1910 with its foundation in Steel, Apeejay Surrendra Group is today an over 6,000 crore enterprise employing close to 43,000 people in rapidly expanding operations in Tea, Hospitality, Shipping, Retail and Real Estate, and has diversified into new business initiatives such as Shipyards, Logistics and Knowledge Parks.

You must have heard about the 'problem of many,' have you? Our project was something to do with problem of 'too many.' In the process of supporting business growth across group companies, we kept on adding individual servers without being able to realise the total return on investment. I am going to share here the story of our 'journey' to server virtualisation, which we started early last year.

We called it Project Shakti – a server virtualisation project implemented at Apeejay. With VMware ESX, the implementation consisted of 12 servers in the database category and 12 in business applications. The remaining servers were distributed among messaging, work flow, gateway and security. Since the last two categories included servers that required intensive management, these were migrated first. Four servers each from the two categories were migrated. Two physical servers from IBM and HP have been used with hard disk capacity of 900 GB each, with 12 GB RAM. These physical servers host 10 and 12 virtual machines respectively. Consolidation of 24 servers has been completed so far.

Now, the important question was, why virtualise? In December 2008, Apeejay's ultra-thin IT team, with just two in-house engineers, started facing several technical challenges in managing the more than 40 individual servers accumulated over a period to meet the needs of the conglomerate. Apart from the manageability, there were many compelling reasons for Apeejay to go down the virtualisation path.

These included resource utilisation; need to implement a scalable IT infrastructure with lowest investment possible; future scalability and support for business growth; and initiating a green policy.  

It was known that virtualisation is the answer to all the above problems but there were two self-imposed conditions. The first was on hardware resource where the decision was to go ahead with the existing hardware and to not buy any new boxes. The 2nd condition was on external resources – the project was to be implemented and managed in-house. These two simple conditions made the implementation a bit complex.

We decided to migrate twenty two servers with a representation from each category using ESX 4.0 and using existing HPDL360 and IBMX3650 boxes with a basic configuration of dual CPU, 900GB HDD and 13GB RAM with VMclient for single window management of the environment. The project has been delivered in three and half months using one in-house resource. In the initial phase, we migrated the most critical ones like 'gateway & security' and business application servers. Balance server got migrated later. The implementation has been designed with complete redundancy between two VMs.

Other than the initial installation support from a local vendor, no external help was taken to implement the solution. However, there were challenges faced during the initial phase of the implementation due to lack of complete technical knowledge on VMware, which was acquired subsequently through self-learning and implementation experience. Though the organisation overall and IT in particular benefited immensely, overall it we felt that a little more aggressive implementation strategy could have yielded better result instead of being defensive. It is better to choose server from each category like security, messaging and business application server in the 1st phase itself. This would have helped in understanding and providing better clarity on which category will work on a virtual server platform and which will not.

As quickness and fast response were the main expectations, we had to adopt a strategy, which was different from a conventional approach of detailed modeling and designing to an agile one, and started off on the basis of overall objective. Support and encouragement from the senior management has worked as an additional advantage.

Apart from the guidance from group CTO, one in-house team member, Senior Manager - IT infrastructure, Joy Bagish and one external resource from the hardware vendor were involved in the complete implementation.

The project has resulted in several benefits for Apeejay in general and Apeejay IT in particular, as follows:

Proactive management of many ( more than 40) of our servers with thin IT resources: After the implementation, there are two VMs configured and currently managed through one VM client instead of individual server management. Resource reallocation is happening on the fly based on the seasonality and statutory requirement of multiple group company businesses under Apeejay Surrendra Group.

Resource utilisation: The utilisation of server CPUs, which were earlier in the range of 15% to 20%, is now 70% to 80%.

To implement a scalable IT infrastructure with lowest investment possible: The complete solution has been implemented by spending Rs. 10 lakhs on software and accessories. Zero cost on hardware.

Future scalability and support for business growth: Three new business servers, one server for HR application, and three IT infrastructure server have been confgured without buying any single hardware and generated savings of Rs. 30 Lakh. A real example of delivering more with less.

Green initiative: Along with allied power and space savings, with an assumption of average 4 tons of CO2 annually per server, Apeejay IT has roughly generated a savings of around 100 tons of CO2 per year. Apart from this, the cost of cooling has come down to one-fifth.

For Apeejay, It’s a journey and a philosophy, which will be pursued and followed up in the coming years as well. The similar setup will be extended to primary DR site at Delhi , Badrapur. Apart from VMWare, Microsoft HyperV are being rolled out in branches at Delhi, Chennai and Mumbai. Apeejay IT promises to deliver an integrated virtualised environment consisting of multiple VM solutions across geographical locations.

By Shubhashis Saha