Predictivie analysis will keep Virtus data centre running smoothly
Data centre provider Virtus is using predictive modelling to improve the efficiency of its forthcoming flagship London data centre – a move which could give better results than traditional tools which analyse recorded results.
The 11.4MW London2 data centre Virtus is building in Hayes, Middlesex, will use all-renewable power, and will also use modelling and analysis tools from Romonet to maintain high efficiency in real time. The data centre also uses indirect evaporative cooling, which Virtus says is a first for a London data centre, and will bring its PUE (efficiency rating) to below 1.2.
Most data centres have problems
“It is disturbing how substantial the problems are with many data centres,” said Liam Newcombe, CTO at Romonet, adding that the answer which is often proposed – data centre infrastructure management (DCIM) is not up to the task of fixing the issue.
“DCIM is entering the trough of disillusionment [in the traditional Gartner hype cycle],” Newcombe told TechWeekEurope at the Data Centre Dynamics event in London last week. DCIM requires a big outlay to install monitors throughout the infrastructure, and then provides too much data, sometimes applying over-complex fluid dynamics to the problem.
“There is little correlation between instrumentation and efficiency,” said Newcombe. “With very little metering deployment, we provide a lot of what people hoped to get from DCIM.” Romonet’s service compares the measured performance with a baseline predicted from the design.
The Hayes data centre will be closer to London and more than twice the size of Virtus existing London1 site, a 4.2MW facility in Enfield. The cooling is also an improvement as Enfield has intelligently managed conventional chillers and a not-quite-so-impressive PUE of 1.5.
Virtus applied the Romonet analysis to the data centre during its design, and will continue to use it in operation, to check that measured efficiency is in line with what was designed. Without this sort of “reality check”, and with “snow blindness” from a flood of data provided by DCIM, data centre owners are often unaware of issues, said Newcombe.
“We’ve modelled hundreds of data centres across the world and the vast majority were not operating as intended. This is no reflection on the designers, engineers or operations people that are involved,” said Zahl Limbuwala CEO of Romonet. “It’s more the case that operators have not had the tools to date to continuously benchmark and evaluate the complex behaviour of their data centers in real-time against the real world conditions.”
Romonet’s service is also now available through the Canara remote power monitoring service, which continuously checks critical power infrastructure. While Canara has already offered remote monitoring, Romonet has added operational and cost modelling to the service.
Do you like renewable energy? Try our quiz!