Computer Laboratory

I am a Research Associate in the NetOS division of the Systems Research Group at the University of Cambridge Computer Laboratory, while I finish writing up my PhD under the advice of Dr Steven Hand.

In the 2013/14 academic year, I was also Director of Studies in Computer Science for St John's College and Peterhouse.

On this page, you can find information about my research and teaching activities. I also have a (rather information-sparse) personal website.

Research

In am interested in the areas of operating systems and large scale distributed systems, and my work often incorporates and combines elements of both.

  • CamSaS: Much of my work is part of the Cambridge Systems at Scale (CamSaS) effort to build a new systems software stack for large-scale data centres, which I co-lead.

  • DIOS: Modern data centres are "warehouse-scale computers", and users expect to see them as such. Yet, our operating systems have not changed in decades. As a result, a data centre machine's OS has no idea that it is part of a larger whole. DIOS is partly a revisit of old distributed operating system ideas in the context of modern data centres and partly a push for more scalable and well-integrated OS abstractions for "data-intensive" computing.
  • Firmament: In "warehouse-scale" data centres, cluster and CPU schedulers are currently entirely separate and share no information. Firmament changes this: like Quincy, it models the entire data centre as a flow network and runs a fast solver to optimize over it. Unlike previous work, Firmament schedules tasks down to individual CPU cores, and deeply integrates local machine information with cluster-level concerns, yielding overall improved placement decisions in heterogenous environments.
  • Omega: Monolithic cluster scheduler architectures have difficulty keeping up with the scale and diversity of modern data centre workloads. Omega adresses the need to cater for this via parallelism, shared state, and lock-free optimistic concurrency control. Omega supports highly heterogeneous schedulers sharing a single cluster's resources without significant interference. This increases scalability and offers the opportunity to rapidly build and deploy new schedulers. This work was done during an internship at Google. [EuroSys 2013 paper]

  • I am also involved in the QJump project on bounded latency data centre networking. [to appear at NSDI 2015]

I also curate our database of inter-process communication performance (IPC) measurements, and run the NetOS group's weekly operating systems reading group, in which we read both seminal papers from the past and modern publications in the general area of OS research.

Once upton a time, I wrote my undergraduate dissertation in Computer Vision, investigating interactive 3D structure-from-motion.

Selected publications

A full list of my publications is available on the publications page.

Supervisions & teaching

For the academic year 2013/2014, I was the acting Director of Studies for Computer Science undergraduates at St John's College and for Easter term, I was also Director of Studies at Peterhouse.

Over the past years, I supervised a range of courses across all years of the Computer Science Tripos, with a focus on core systems courses.

In addition, I have supervised a number of undergraduate final year (Part II) projects, including three award-winning ones, and co-advised several masters projects.

Contact me

Email:

Office phone: +44 1223 763683

Snail mail:
Malte Schwarzkopf
University of Cambridge Computer Laboratory
Willam Gates Building
15 JJ Thomson Avenue
Cambridge CB3 0FD, UK