Real problem, virtual answer

June 21, 2005
This article originally appeared in DigitalLife:

Abacus International's old IT infrastructure could no longer meet the demands of its solutions group. So they switched to a new virtualisation platform - and saved a million a year on IT costs. CHAN CHI-LOONG reports

Wires everywhere. More than a hundred PCs, servers and printers stacked in unsightly tangles across four rooms. Worse, a power trip once or so every four months, which shut down critical services. To top that off, Abacus - a global travel booking company headquartered in Singapore - needed more computing power and resources to support its growing solutions business. And the company was fast running out of office space.

With 11,000 travel agencies in 22 markets in the Asia-Pacific region as customers, Abacus needed the computing power to service all of its clients' computer reservations systems.

Additionally, it also needed to support a host of solutions it rolled out for some of these agencies - like billing, point of sales and accounting.

Whenever a customer called up with a problem, the support team had to replicate the client's server environment in order to debug it.

Without automated test tools, everything was done manually, and this contributed hugely to the server sprawl.

Fixing the problem

'We needed to fix the problem rather than wiring more cables o moving more machines about,' said Mr Lim Lai Hock, deputy directo of solutions at Abacus.

The company tried standardising configured hardware and using Ghost, a popular hard disk imaging solution. But that wasn't the answer.

For one thing, it still didn't solve their server sprawl woes. Also, it did not reduce the turnaround time for customer support calls by much.

So in 2003, Abacus approached its long-term systems integrato partner, JOS Technologies, for a recommendation.

Virtualisation was to be the way out. (See Virtualisation?). Afte three months of testing on prototypes, Abacus finally selected VMWare as the virtualisation platform of choice.

Just four IBM servers running VMWare's solutions replaced all of the previous 60 servers - and that from a mishmash of vendors like Compaq, HP and Dell. Next, 24 new workstations replaced the 70 or so PCs being used.

Lastly, all the new equipment was packed into one newly renovated server room, freeing up three rooms.

The cost of change: $400,000. Of that, $270,000 went to the new servers alone. The rest of the cost went to the workstations and the renovation of the new server room. Despite the price tag, Mr Lim is happy with the results. He estimates that the new VMWare virtualisation platform has saved Abacus a million dollars a year.


Turnaround time has been shaved to half a day per problem - from an average of one-and-a-half days, a more than 50 per cent productivity increase.

That's because technicians do not have to run around and rewire and reconfigure the PCs and servers.

Instead, they simply request resources from the platform, and a virtual machine, which replicates the client's environment, is created for them to work with. Testers can access this virtual machine remotely, too.

Another perk: Abacus pays less for maintenance, power consumption and manpower. For example, the solutions division has shrunk its headcount to 67 now from 110 in 2002, despite a heavier workload.

With the new platform, the solutions division can try new projects it could not possibly do before. One example Mr Lim mentioned is stress testing their applications by simulating heavy load - creating hundreds of virtual machines on their virtualisation platform to connect to their applications - to measure how they will perform.

The benefits have even percolated to the staff at the lower levels who work in the new environment. Said Mr Yeo Siow Keong, senior support engineer at Abacus: 'It is great because now I can manage the entire infrastructure from a single view. No longer do I have to walk around to configure individual servers or carry heavy servers!'


In virtualisation, physical computing resources like storage space, CPU power and memory are mapped and abstracted behind a logical structure.

This means, for example, that a network administrator doesn't need to care which physical machine holds his data, but requests space directly from a pool of available resources.

In this way, servers can be consolidated to one logical pool of resources, and managing the resources is made much easier.