Last week, I highlighted some points from Professor Mor Harchol-Balter’s talk. This week, I would like to focus on a different point she made related to the academia-industry divide.
If you have read Mor Harchol-Balter‘s CV, you may have noted that she started off in a more theory-oriented area of computer science. Her recent work related to server farms, while grounded with theoretical foundations, has ventured solidly into empirical measurement as well. This has proven to have its upsides and downsides.
On the downside, this has made publications trickier. Most publication venues have a bias towards either theoretical or applied/empirical work. Publishing in most venues thus requires tweaking papers to take on a more theory-based or systems-oriented angle. For someone as versatile as Mor, this probably isn’t a huge problem, though it does inevitably slow down the pace of research.
On the positive side is relations with industry. Large businesses, rightly or wrongly, are distrusting of completely theoretical results. On the other hand, projecting computing requirements to the orders of magnitude required by the likes of Internet giants Akamai and Amazon is difficult based on empirical research done on a much smaller scale in universities. Mor’s work bridges the divide by providing a theoretical basis that accurately models results obtained.
Perhaps academics have something to learn from industry in this regard. By combining theory and application, theories are more readily falsified. Meanwhile, a better understanding of real-world processes through theory is in itself beneficial both from an intellectual standpoint and from a practical standpoint. This mixed approach should appeal to scientists because it is a form of replication: two very different methods are used to arrive at a conclusion and, if they are consistent, our confidence in them should be bolstered. East is east, and west is west, but perhaps one day, theoreticians and empiricists shall meet.