Travel data centre serves up speedy requests

Pegasus Solutions has found a cost-effective and energy-efficient way to handle the more than one billion travel and hotel transactions processed at its data centre each month.

Pegasus handles back-end transactions for online hotel and travel reservations for clients such as the Savoy, Thistle and Travelodge. It must respond to real-time requests for travel information - and as more and more people arrange travel plans online the number of requests has skyrocketed.

Pegasus CTO Steve Lapekas said it's not unusual for a person to conduct 50 or more searches for each single booking they place. This means Pegasus' data centre is swamped with requests for fresh data which cannot be cached - and the centre needs to be able to process these requests as quickly and accurately as possible.

"What we do at Pegasus is not overly complicated but it's in large volumes," he explained. "Every day we don't know what the levels [of activity] will be."

To speed processing at Pegasus' data centre, Lapekas chose to install hardware appliances from Azul Systems. The centre uses two 11-unit rack-mounted systems with 192 CPUs and 64GB RAM each for production, and an additional system with 96 CPUs and 64GB RAM for development and testing.

Azul's processing technology and hardware will save Pegasus $70,000 over three years on power alone and Lapekas estimates the company spent half as much on the Azul hardware as it would have on comparable hardware with the same raw computing power.

The power savings were a big win for Arizona-based Pegasus which must pay considerable energy bills to cool its data centre.

Lapekas admits that given how large the data centre is, the amount saved on power is relatively small but added: "Every little bit helps."

With the travel industry growing and more and more people booking travel online, the demands on Pegasus' data centre will only grow. But with the Azul hardware in operation for nearly a year, Lapekas feels confident he's getting the fastest processing he can for the money he's invested.

"A data centre is [a] math equation," he said. "At a square foot level, I believe I've optimised as much as I can."