Motivated by massiveness
Alexandru Iosup, full professor and chair of the Massivizing Computer Systems group at VU Amsterdam, likes to pursue great dreams. ‘I am on a quest for a grand unifying theory that explains how computer ecosystems work, with the aim to improve their performance and make an impact on millions of people.’
How did you end up in computer science?
‘I was a curious child, with a mind that liked to logically structure what I observed in the world. When I was growing up in Romania, computers were rare things; I saw my first computer in a dedicated lab in Bucharest when I was about 8 years old. Since computers embodied a world of structure that I instinctively understood, I was immediately intrigued.
When in high school, I followed a dedicated track on informatics, but it was too basic to fit my interests. So, when I was about 14, I took a professional course in C++ programming, where I was that one teen among many people who were all over 40. As soon as I understood that I could make a living out of programming, I wanted to quit school. Luckily, my parents convinced me that wouldn’t be such a good idea. So, I stayed in school, and in my spare time, I worked as a programmer, mainly for companies that developed games.
At that time, the Internet started to become a thing. Since I did not know anything about networks, I enrolled in a computer science program. But I did not do so with the intention of becoming an academic, since I was oblivious to what science actually is. Later, my decision to pursue a PhD was motivated by my wish to understand distributed systems, still thinking that would be useful if I were to become a game creator. However, I got excited about not only understanding how things work, but also about explaining why they work that way. Why do computing systems show different behaviour from what we would expect from theory? Is there any grand unifying theory that describes how computing systems work, and that can help us improve current and future systems? Pursuing that theory has been my main motivation in my scientific career ever since.’
What is your research about?
‘In our current group at VU Amsterdam, we work on Massivizing Computer Systems: we look at computer ecosystems consisting of multiple types of hardware and software, made and integrated by different organisations, that together perform some function, but they also do it well. Think for example of datacenters, cloud computing, and big data and AI applications. In our research, we create theories about their operation and use these to design, develop, deploy, evaluate, and benchmark useful distributed computing systems and ecosystems.’
Could you give an example of such a system?
‘Take the Dutch railway company NS. The brains of NS are formed by a myriad of computing systems, that operate together to perform some function. When I check-in here in Delft and check-out in Amsterdam, there is a system that reads my card, there is software that figures out where and when I am traveling, and some other system sends this information to some banking computer that has to process my payment. All of these systems have been created by different companies for different purposes, and are based on different ideal hardware and software platforms. In our research, we take a holistic perspective on such collections of computer systems: we look at them on different resource-scales and timescales, ranging from their individual components and microseconds to the entire system and timescales up to years or even decades. We try to explain what is happening at each level, and also unify the understanding across the entire ecosystem. The core question is: can we explain how computer particles get to be put together into a larger ecosystem and how such a system can be operated as efficiently as possible in terms of energy usage, data input, computing times, and so on.’
Since these types of massive computer systems are omnipresent in our current day and age, how do you decide which topics to work on?
‘When I started this group back in 2018, together with the team I developed a very clear mission and vision: if a proposed research subject has the potential of reaching millions of people within the foreseeable future, then we’re in. We want our results to end up as the de facto standard in the community.
Since we cannot compete with big tech on the engineering side, due to sheer scale, instead we identify topics when they emerge and provide scientific insights. We for example identified the potential of server-less computing at a very early stage: that development started in the industry in 2016, and our first position paper about it came out early 2017.
Another thing I am passionate about, is data collection for the wider computer science community to profit from. We for example pioneered collecting long-term evidence about the performance of various data centers, recording all information that was available, like workloads, scheduling schemes, types of hardware, et cetera. You can use these openly available archives to build queries upon and validate individual scientific theories, and in practice to test and benchmark designs to full production systems. From these data, we have for example found that the many of reported performance results are not reproducible. For example, if you present a certain job to a specific data-center, and you present the same data-center with the same job at a different time, the processing time may vary between, say, 2 seconds and 25 minutes, making statements such as “10 percent faster” a statistical game rather than a useful fact.’
What is it that fascinates you in this field?
‘The ecosystem idea changes the game in computer science. Such a system consists of computing units that are not built with the same clear goal, not by you, not for your purpose, and often you cannot influence them. It is this complexity that makes it difficult, yet intellectually extremely challenging, and highly rewarding work.
Computer systems are the backbone of our society. That means we have to manage them properly. There must be access to a service whenever you need it. Imagine having some digital lock on your door that scans your face to provide you with access. If the database containing your picture for reference is down, or some other of possibly hundreds of services that all must work to check your home-access is down, you are effectively locked out of your own home, even if the failure is not on your side. Similarly, large-scale societal infrastructures like hospitals or railways can be down for hours or longer, as a result of computer systems problems they cannot fix themselves.
To help, in a new, large-scale program, we are now building a digital twin of our digital infrastructure to mimic real world ICT networks. The aim is first to understand what is happening now and then run scenarios to inform stakeholders about possible issues including failures, and how to circumvent them. For example: “We anticipate the network to suffer then and then, so back up your files now.” Or: “Deploy your operations elsewhere to cut energy usage by fifty percent.”’
You act as a co-chair of IPN’s Special Interest Group Future Computer Systems and Networking. What is the ambition of this community?
‘I want to thank professors Fernando Kuipers, Paola Grosso, and Ana Lucia Varbanescu for co-leading this effort, and the whole community for their collaboration. After about a year of intensive collaborative work, in 2022, we issued a manifesto summarising our main ambition to create a Netherlands community of computer systems and networking researchers, both in academia and institutions with links or interest to research. Together with many partners, we literally put the community on the map. We aim to operate long-term at the top level of scientific research, worldwide. We’ve set a clear vision to communicate our strengths and explain our added value to national and international stakeholders. Also, we want to organise research around this vision to be more effective and efficient with available funds and resources. We have already attracted important funding for education and a Growth Fund program for collective R&D&I work. We are currently contemplating setting up a dedicated national research infrastructure, and we are discussing many topical working groups and a national education agenda as well.
As a community, we are actively reaching out to potential collaborators outside our field. We urge them to make use of our knowledge and expertise to let us help them help the Netherlands move forward and become more digitally sovereign in this era of digitalisation.’
Foto: Sjoerd van der Hucht