HP Extreme scales out beyond blades

The company said that the new portfolio would allow organisations that operate thousands of servers to scale extremely heavy workload requirements to a lightweight infrastructure more quickly and cheaply.

Based around HP's ProLiant SL server family, the ExSO portfolio has been designed to include a modular systems architecture that replaces traditional chassis and rack form factors with a rail and tray design.

Michelle Bailey, a research vice president at analyst firm IDC, explained that businesses built on extreme scale-out environments, such as cloud computing, Web 2.0 and high performance computing, operate at maximum transaction volume and low margins.

"These customers have very distinct and unique datacentre requirements, specifically around energy efficiency, cost and time to market," she said.

HP claimed that ExSO has been designed to offer a consolidated power and cooling infrastructure that uses 28 per cent less power per server than traditional rack-based servers.

At the same time, reductions in the metal used in the ProLiant SL servers have reduced weight to help lower shipping and floor support costs.

Modular configurations with swappable 'compute trays' have been designed to allow rapid deployment, while enabling double the density of traditional racks to deliver a single node per 1U chassis, with up to 672 processor cores and 10TB of capacity per 42U rack.

New HP Datacenter Environmental Edge software, meanwhile, has been introduced to enable visual mapping of environmental variables, so that datacentre managers can quickly identify and take action on performance, power and space inefficiencies.

"There is certainly a trend towards scale-out architectures, and Cisco did something similar recently," said Dale Vile, research director at analyst firm Freeform Dynamics. "But the challenge is that most corporate datacentres run a mixed set of workloads."

Vile added that, while many companies are still virtualising and standardising datacentre infrastructures, it will take a further 12 months to establish which homogeneous workloads using standard processes, like email for example, will best suit such scale-out environments.