The aim of AutoDC is to provide an innovative design framework for autonomous data centers.
Autonomous data centers should be able to, without any human intervention, from a best effort perspective continue its operation and self-heal independent of contextual interference, such as intermittent power failure or overheating to name a few.
The usage scenarios range from large to mega-scale data centers being colocation, wholesale or web, edge data centers, remote data centers and data centers in developing countries.
With an expected continued growth in the data center market, the cost of operating and maintaining the data center footprint will increase. The administration and maintenance cost in large and mega-scale data centers is one third of the OPEX and challenges in this market segment are due to the vast amount of equipment and personnel.
Part of the growth, except in the usage scenario of mega-scale data centers, is predicted to be in the market segment of edge computing, where infrastructure is close to application usage such as urban areas, intra-urban transportation routes and areas of dense congregation of devices.
These future requirements will need distributed data centers and equipment leading to an increase in the cost of operation and maintenance.
The project will specify and design data centers to be fully autonomous, with no requirements for on-site maintenance, and preferably minimum remote maintenance. Beyond the obvious need for adding more redundancy in the form of back-up components, this level of autonomy will require a detailed knowledge of the ecosystem’s condition and context. It also requires beyond state-of-the-art ability to autonomously detect failures and maintenance activities as well as controlling the environment employing AI trained software, without human intervention.
In addition to significant OPEX reductions, due to lower maintenance and operation costs, autonomous data centers also become key enablers of markets in developing countries, where the lack of on-site knowledge and staff inhibits their journey towards digitalization. Another interesting use-case and market is remotely located data centers, which have high maintenance costs.
Fulfilling the vision of completely autonomous data centers requires many technologies to be combined in the technical value chain. Data center complexity is increasing and there is a need for holistic and integrated approaches. Developments in modular data center components lend themselves to applications, and will facilitate this project by enabling experimentation to take place in laboratory environments. However, the concepts developed will be clearly scalable to larger applications.
Sensors measuring the physical properties of components and their environment in addition to equipment that monitors power and cooling must be well integrated and improved from today’s state-of-the-art offerings. Innovative and sophisticated control peripherals will also be required to
replace human intervention.
A powerful data analytics engine is required to achieve data collection from the various monitoring systems, which is then consolidated with external data sources and periodically stored as appropriate records to allow for both real-time and off-line ecosystem modelling and machine
learning data analysis. The analytics results will ensure proper actions are applied to the control systems for optimised power, cooling, network and server operation, which is essential to maintain the data center “health” within desired parameters to reach identified target KPI values.
Expected project outcomes and dissemination:
The consortium consists of partners from Europe (SE, FI) and Canada, with Ericsson, ABB, Granlund and Swegon being the major industry partners. Several relevant SMEs, including Swedish Modules, OP5, Hi5, Clavister, Comsys, SEECooling, Missing Link Technologies, Mariner partners, ionSign, OuluDC, HitSeed, Cumucore and SenSoftia bring valuable hands-on experience. A group of respected research organizations and universities provide academic excellence in several fields of research to achieve this novel vision of autonomous data centers.
By the end of the project our aim is to have a reference design for an autonomous datacenter ready to for construction, possibly also a proof-of-concept during project. Along the project, we will test and demonstrate important technology steps using the RISE SICS North datacenter lab SICS
ICE (http://ice.sics.se). By using this laboratory we can do full scale test with up to 100kW load in a safe environment. Simulation of power outage, network interference and other possible failures can easily be simulated without tampering on security and data integrity, which would be the case if using a commercial datacenter. Other dissemination activities include demos and presentation at datacenter conferences, publication in scientific journals and conferences.