The ECORAN project is reducing power consumption by introducing novel ways of interconnecting and managing servers, accelerators, storage and interfaces in small processing cells (cloudlets)
Total project funding amount: £492,000
Project date(s):
Category: Government Funded
Location(s): Leeds
Status: Open
Partners
University of Leeds, Ultracell Networks Ltd
A key feature of ORAN is the use of commodity hardware made up of servers, accelerators, storage and programmable interfaces, together with open software.
The ECORAN project is reducing the power consumption of the commodity hardware used by introducing novel ways of interconnecting and managing servers, accelerators, storage and interfaces in small processing cells (cloudlets).
The open and intelligent processing on demand approach can enable a wide range of applications and can reduce the total cost of ownership significantly. For example, demand peaks may be witnessed in stadiums, in shopping centres and in city centres for short durations during a match, during the lunch hours or during an event. Processing on demand enables the correct amount of processing to be orchestrated and provisioned on demand at the correct location at low power consumption and low latency using ORAN.
Project Overview from Innovation Briefing Issue 8
The Energy-efficient Cloudlets for ORAN, or ECORAN project will reduce the power consumption of the commodity hardware used by Open RAN.
A cloudlet is a small-scale server cluster which provides low latency data processing close to the smartphones, tablets and wearable devices it is designed to serve. It is a form of Edge processing. While Fog Computing needs connection to the cloud and Mobile Edge Computing is always standalone, Cloudlets can operate in either mode.
ECORAN looks to reduce the energy consumption of cloudlets by introducing novel ways of interconnecting and managing the cloudlet servers, accelerators, storage and interfaces.
The project is led by the University of Leeds in partnership with Ultracell Networks.
ECORAN takes an intelligent approach to scaling capacity as it is needed. For example, demand peaks may be witnessed in stadiums, in shopping centres and in city centres for short durations during a match, during the lunch hours or during an event.
The servers currently consume around 20 per cent of the total power consumption of the Cloudlet. This is expected to rise to over 80 per cent of the total power consumption. The processing capability per server has increased slowly, but the network interconnecting servers has evolved from 1 Gb/s to 10Gb/s server interfaces, to its current standard 100Gb/s rate. Speeds of 400Gb/s per server connection are expected soon. Initial results show that the power consumption of the network interconnecting the servers can be reduced by around 82 per cent and a similar reduction in latency, of around 85 per cent, is observed by eliminating the hierarchies of the switches that connect the servers.