When Japanese investment bank Nomura acquired the Asian and European businesses of Lehman Brothers as it collapsed in 2008, it suddenly needed to expand its data centre capacity to support the new divisions.

At the time, the bank’s own facilities were approaching maximum capacity, so in early 2009 it decided to move to two new, colocated facilities outside the M25. It procured and installed 3,500 new servers, and in May 2010 migrated across the Lehmans’ applications (it is completing migration of its own software infrastructure this month).

Unlike normal data centre migration projects, which have long timescales, moving Lehman’s infrastructure into the new facility had very short, and very definite, deadline. This meant that the priority during migration was to get systems up and running as fast as possible.

There was not time, however, to make sure the allocation of power to hardware resources was perfectly optimised. After the switch over, therefore, some of the power allocated to the server racks was not being used, imparing the efficiency of the facility.

According to Mark Andrews, a data centre manager at Nomura, one of the first jobs following the initial migration was therefore to build an accurate picture of the data centre’s inventory and how it relates to power infrastructure.

For the first part of that, Nomura used its self-built asset relationship management (ARM) tool. This is essentially a database of equipment, with links to the applications and systems it supports.

Andrews and his team first collected data describing the new hardware. In one of the data centres, after the hardware manufacturer failed to supply this data as expected, that meant physically walking through the aisles scanning serial numbers and writing down details of each asset.

Modelling power

To model the power infrastructure, Nomura adopted data centre power management software from nlyte. This allowed it to build a model of the data centre infrastructure, including both IT equipment and power infrastructure, from the power distribution unit (PDU) down to the plugs.

Integrating the nlyte software with the ARM tool and systems that monitor the performance of energy infrastructure and IT equipment in real time allows Nomura to see which hardware is maxing out its allocation of power, and which is not using its total allocation.

This is has allowed it to reorganise the layout the data centre to ensure maximum power utlisation. "We try to make sure every cabinet is using all of its power allocation," explains Andrews. "It’s about maxmising the use of the power that has been allocated."

Adding data from Nomura’s environmental monitoring tools within the data centre allows it to see which equipment is producing the most heat, says Andrews. “We can see where the hot spots are in the data centre, and we use that in discussions with the landlord about the ambient air temperature.”

The nylte deployment may have implications for the way in which Nomura’s IT department charges the business for its resources. Currently, infrastructure costs are charged on the basis of hardware. But in future, Andrews says, it could use data from the nlyte tool and the ARM system to charge business units on the basis of consumption – both of IT resources and electricity.

“We’ll be able to say, this server is drawing X amount of power and costing us X number of pounds a day to run,” he says. “By linking that back to the application owner, we could charge it back to the business. We’ve already got all the information we need.”