RISE SICS North Collaboration, Sweden



Hardware v Software

To date, the datacentre industry has been almost entirely focused on hardware improvements to gain efficiency - more efficient chillers, servers, rack layouts, etc.

As a result, the industry has made major strides in improving efficiency – the industry standard measure of PUE (Power Usage Efficiency) has been steadily moving towards the magic figure of 1.0 but further gains are becoming more difficult and expensive to achieve.

What about the software?

Despite the primary purpose of a datacentre being to run software, little has been done to utilise software-based methods to improve efficiency, whether that is reducing power consumption or increasing throughput. 

The reason is that datacentres are very complex combinations of many components that behave in hard-to-predict manners - each datacentre is unique in its overall characteristics. Software that can improve efficiency must have real-time insight and understanding of the behaviour of all the elements in a datacentre – a major challenge, and beyond the reach of most operators and integrators.

Software represents a new opportunity for efficiency gains that can extend and augment the gains that have been made to date using hardware.

Edgetic’s advanced software and machine learning methods deliver automatic, self-managing technology that monitors the software environment of a datacentre and optimises the provisioning of workloads across the servers to gain efficiencies in power consumption and throughput.

The Edgetic approach

Edgetic is a fully automated software framework that can be smoothly deployed onto any container based (“serverless computing”) cluster, big or small. Edgetic analyses datacentre metrics collected in real-time, including temperatures, CPU loadings, disk activity and network traffic, as well as a variety of advanced readings collected through our advanced software monitoring technology. Unlike infrastructure management products (e.g. DCIM), Edgetic is non-intrusive and simple to implement.

Using state-of-the-art machine learning, behavioural models are built and refined to enable automatic, real-time decisions to be made and carried out, ensuring optimum performance at any moment. Along with predictive behaviour models, the Edgetic approach enables continuous ‘learning’ and adaption to the target datacentre, its characteristics and the demands of its client workloads.

Edgetic can also provide advanced anomaly detection; if the behaviour of any aspect of the system (software, hardware, or even the thermal environment) differs significantly from predicted models, alerting human operators to check for malfunctions or even a malicious event in the system.

Finally, Edgetic’s advanced behavioural models offer state-of-the-art capacity planning tools. Since models are automatically tailored to each customer system, simulation-based modes enable advanced “what if” decision support based on accurate predictions. For example, what is the impact on my system if my biggest customer doubles their demand? What will be the performance impact to my customers if I reduced my server count by 200?

What does this mean for a Datacentre Operator?

Edgetic offers a software-only means to improve the efficiency of a datacentre, whether it’s a small, server room or a larger-scale centre with thousands of servers.

Edgetic can deliver substantial reductions in power consumption without the need for any additional capital expenditure or changes to existing hardware systems. Edgetic can improve the throughput of each server within a datacentre – ensuring optimal software performance per watt and increasing the level of effective utilisation.

Edgetic provides these capabilities as a fully automated and continually adapting service with minimal overhead, implementation or configuration.

Whether it’s for cost savings, pursuit of green credentials or increasing job throughput, Edgetic can increase the financial performance of a datacentre and deliver a rapid and easy Return on Investment.