Introduction

It is estimated that the energy consumed by data centers – used to power both the servers themselves and the cooling systems – is equal in the United States alone to 120 billion kWh, corresponding to approximately 3% of the total electricity produced by the country. Energy consumption impacts both on the costs of the companies that own the data centers and on their customers, as well as on the environment due to high CO2 emissions. Eco4Cloud is a solution for modernising corporate data centers and exploiting the potential of Cloud Computing, to obtain greater efficiency, minimum guarantees of quality of service for applications running on data centers, and significant savings in energy consumption.

Technical features

The power consumption of data centers is related to the fact that servers are used only for 20-30% of their capacity, and a server with low load consumes about 70% of the energy consumed by the same server at full load. Current solutions to this problem are based on a decentralized approach that seeks to optimize the distribution of applications on the various servers through a central manager constantly updated on the status of the machines, the concurrent migration of many applications which, however, often generates a decline in performance, the all without reaching the optimal solution. The patent is based on an innovative bio probabilistic algorithm, which enables energy, cost and carbon emissions reductions of up to 60%, as well as effective risk monitoring, planning, optimization and mitigation. The approach used is decentralized / statistical and requires each server to decide independently whether to accept an application or whether to request migration to another server as the workload changes in real time. The use of Eco4Cloud can halve the energy consumption of a Datacenter. TRL: 9 (http://www.eco4cloud.com/).

Possible Applications

• Can be used in medium-large data centers (from a few tens to thousands of servers).

Advantages

• Decentralized, scalable, and fault-tolerant management;
• Migration of many applications is gradual and continuous, to avoid possible performance decay;
• Statistical approach ensures the minimum energy consumption, especially in data centers with tens or hundreds of machines;
• Real­time monitoring and analytics on multi­dimensional utilisation of critical data center resources.