Confessions of a Virtual Transplant

BANGALORE, INDIA: Sujyot Talaulicar is Manager IT and Systems at Vertex Software Pvt ltd which is a company established in 1996, in the playfield called software testing and development, mobile technology development, etc.

When faced with challenges like unsustainable power and cooling costs, Vertex decided to give a thought on virtualization, and here’s what unfolded for the man in the driver’s seat.

Why virtualization and why VMware?

We have multiple projects in the IT action-field. Each project requires their own test data centre to be built till we consolidated it to VMware. We had high-end desktop environments with heterogeneous platforms. We found it tough as hundred kinds of platforms are difficult to manage. The hardware spread included three floors across three premises. When we conducted application testing and staging on 100 desktop-class machines deployed across three floors, we found its management load and power and cooling costs were becoming unsustainable. After reviewing virtualization products from rival vendors, Vertex opted to deploy an infrastructure based on VMware vSphere 4. The company engaged VMware Partner Sunfire Technologies Pvt Ltd to conduct the implementation.

How did you transcend the usual obvious worries around visibility and security when one has to switch from physical servers to virtual ones?

That actually depends on an organization’s policy and how you implement the new set-up. It’s about how administrative powers are distributed and allocated. Since we have segregated responsibilities between one or two key personnel, it’s taken care of.

Why not other equally strong players in the market? In terms of your evaluation approach?

I attended seminars of almost all of them and evaluation angle was around client-experiences, ROI etc. We conducted a POC and checked all areas an all scenarios with VMware. Other options were definitely explored and evaluated. The reason we didn’t go for players like Microsoft or Citrix because we didn’t see enough implementation done so far. It was also about the kind of environment we were looking for. It was a critical environment with huge investments and with VMware’s support line-up, we were inclined towards them.

Still, any gaps or suggestions for them as per your experience?

As of now, the test is meeting our expectations. To further dive deep into the service aspects, we are doing a POC on desktops also. Support is where the company can improve on strengthening a very good domain.

What’s your take on the argument of one hypervisor vs mix of hypervisors?

I think that if one vendor offers good support, it’s better to go for that instead of investing in multiple vendors.

Do you like ideas like Cloud Manager or Self-provisioning ware, as being dished out by some well-known Virtualization players?

Self provisioning is what we do ourselves. We create an image and give it to hardware, desktop teams. It’s not for the users. They see the same desktop environment. Backstage is not what we worry them with. Depending on scale, scope etc, automization can be allowed, but again on a case to case basis.

What is the status change now?

Vertex is now running an infrastructure comprising about 100 virtual machines running on five host servers connected to a HP StorageWorks Modular Smart Array 2000 storage system. It has started extending virtualization into its production environment, with one intranet and one proxy server virtualized by early January 2010.

 

How was the migration experience?

Pretty good, we got a good team. No hiccups. There were some problems here and there initially around language and non-engineering flavors but then it was a smooth experience.

If you were to list down some major benefits that you have managed so far, what would those be?

Well, for one we have achieved a 20:1 server consolidation ratio and plan a 25:1 consolidation ratio. Plus, we have retained plenty of headroom by achieving CPU utilization of under 50 per cent on host servers. Not to forget the reduced time to deploy new machines from two days to less than four hours.

Specifically, in terms of IT management and the pain points that it entails?

Top of the mind aspects would be maintenance of devices, realty costs, downtime, power costs, project delivery time and redundancy issues. We lowered time required to manage the IT environment by 20 per cent and enabled IT team members to devote more time to delivering business projects. We cut power usage to one-fifth of previous requirement; slashed real estate requirement from three datacenters to one rack in one datacenter, enabling the reclaimed space to be used by employees; and eliminated unplanned outages that extended the time required to deliver applications to customers.