More and more attempts are being made to apply new methods of artificial intelligence to servers in order to use them in an increasingly profitable way. In this regard, a new system developed by computer engineers at the Massachusetts Institute of Technology helps analysts and server administrators to plan data processing activities and does so for thousands of servers.
Currently, algorithms help to develop such plans, algorithms that are often inaccurate and at least man-made. Using a machine learning system such as this could instead help, as described in the press release on the MIT website, “today’s energy-hungry data centers are working much more efficiently.
Indeed, the MIT system provides for the manual encoding of the algorithms by computers. This is done thanks to the technique defined as “enhanced learning”, a machine learning technique “to adapt planning decisions to specific workloads in specific server clusters”.
Even during the “training phase” of the software, no human intervention is required, except for simple initial instructions. Researchers calculate that this new system is 20 to 30% faster compared to the best programming algorithms specifically designed for servers and encrypted by hand, thus saving time and energy.
Mohammad Alizadeh, one of the authors of the research and researcher at the Computer Science and Artificial Intelligence Laboratory (CSAIL) at the University of California, comments on the usefulness of this new system based on artificial intelligence: “There is no solution for making planning decisions. In existing systems, these are coded parameters that you must determine in advance. Instead, our system learns to optimize the characteristics of its planning criteria, depending on the data center and workload.”
Latest posts by Janice Walker (see all)
- Scientists discover that abundant minerals are proton conductors - November 2, 2019
- Scientists sequester the duckweed genome and discover genes that protect them from parasites - October 22, 2019
- Artificial intelligence learns complex patterns of genetic diseases - October 12, 2019