How would you go about organising a time-limited double round-robin tournament involving 16 to 20 teams, taking into account the diverse and occasionally competing priorities of clubs, venues, sponsors, broadcasters, law enforcement, and fans?
“These problems seem really simple,” says Dr Carlos Lamas-Fernández, University of Southampton academic and joint winner of ICT2021, a leading international competition on sports timetabling. “You just get pen and paper and decide who’s going to play against who.”
But the reality, explains Dr Lamas-Fernández, is far more complex. “The number of solutions that are possible is in the billions,” he says, before swiftly correcting himself. “Trillions.”
It more or less goes without saying that generating these innumerable solutions would have been practically impossible without high-performance computing—in this case, the University of Southampton’s powerful Iridis 5 HPC cluster.
Background
The knotty computational challenge of sports timetabling has been the subject of academic inquiry since the 1970s, growing into a significant research field over the last four and a half decades.
However, much of this research is subject to confidentiality agreements, and designed to solve specific problems with unique parameters. Despite many years of effort, there remains no generally applicable method for automated sports timetabling, resulting in countless collective hours spent solving the same problems over and over again.
The International Timetabling Competition is an open research initiative designed to resolve this chronic inefficiency.
It encourages researchers to develop new tools and methods in pursuit of the best possible solution to a common set challenge, consisting of 45 problems across three tranches of instances—novel computational conundrums that competitors must solve in early, middle, and late stages of the competition, under increasing time pressure.
The victorious 2021 University of Southampton team comprised the aforementioned Dr Carlos Lamas-Fernández, Lecturer in Business Analytics and Management Science, and two others: Dr Antonio Martinez-Sykora, an Associate Professor with a research interest in Combinatorial Optimisation; and Professor Chris Potts, a highly experienced academic with a background in the same field.
All three are members of the University of Southampton’s CORMSIS (Centre for Operational Research, Management Sciences and Information Systems), one of the largest and most influential groups in the UK concerned with mathematical sciences and business studies.
How HPC Helped
In order to carry out their research, the team relied on the University of Southampton’s powerful Iridis 5 HPC cluster.
“The main purpose of using the cluster,” says Dr Martinez-Sykora, “is to have different threads looking at different options.” These different threads allowed the team to generate solutions much more quickly and comprehensively than would have been possible on a desktop computer.
At one point in the process, the team tasked Iridis 5 with solving 500 problems in around 40 hours. By comparison, a lone desktop would have required well over two years to perform the same calculations, and throughout that time this industrious machine would have been unable to assist with anything else.
The processing bandwidth of the Iridis 5 HPC cluster not only allowed the University of Southampton team to work through a greater range of problems, but it enabled them to yield more and better solutions to those problems.
“The longer you run it, you keep improving the solutions to get to something better and better,” says Dr Lamas-Fernández.
Eventually, in the judgement of ITC2021, they found the best solution.
Accessing HPC
Dr Lamas-Fernández, Dr Martinez-Sykora, and Professor Potts all have backgrounds in applied mathematics. Their experience in coding made working with Iridis 5 a fairly straightforward process.
A coding background is by no means necessary, however. While previous versions of the software required a sound working knowledge of Linux, the arrival of click and drop interfaces and other user-friendly front-end improvements have brought HPC within reach of researchers who lack coding experience, as long as they’re prepared to learn a few Linux commands.
But the coding demands of a project naturally scale with its complexity. A relatively inexperienced coder can induce Iridis 5 to run a task many times in tandem (this type of computation is called “embarrassingly parallel”), but even seasoned coders like Dr Lamas-Fernández and Dr Martinez-Sykora admit that they would need additional support to develop codes capable of allowing these parallel computational strands to interact with one another.
This support is available at the University of Southampton:, As part of the University’s ongoing commitment to High Performance Computing (HPC), three new specialist Research Software Engineers (RSEs) have been employed to help researchers make the best use of the world class facilities here at Southampton.
For more information, including details on how to get help from the HPC
RSEs free of charge, see: https://rsg.southampton.ac.uk/hpc