LunCH Lecture – Computer Science by Georgios Andreadis
Every day, we all use multiple datacenters. Paying at a restaurant? Your payment travels through a datacenter. Checking in at the train station? Datacenter. Watching a movie at home? Another datacenter. To ensure that all these cloud services don’t fail, the datacenters behind them are often equipped with lots of additional computers just in case something goes wrong. This has a severe price, though. Datacenters already consume a lot of energy and adding more computers is just not sustainable. So how do you strike the balance between too few and too many computers? This is exactly the challenge of capacity planning. For physical infrastructure, we know how to capacity plan: We calculate and simulate what could happen in the future, with the help of computer models. For datacenters, however, we unfortunately don’t have these kinds of tools, with as consequence that industry often over-estimates the computer power needed.
Georgios Andreadis, Best Graduate of TU Delft 2020, developed a tool to address this challenge, in his MSc thesis at TU Delft. Attend this lunch lecture to learn more about how we can make datacenters more sustainable while still resilient to failure. Georgios will also give an insight into his current work on medical image analysis for cancer treatment as a PhD candidate at the LUMC and CWI.Register now