LUMS | LUMS CS Dept.
Home |

Systems Research

Although I like to think of myself as a Computer Vision researcher; over the last year I have been somehow pulled into this ever-enticing field which is plainly called "Systems". I feel that every Computer Scientist should step into this arena at least once, as it gives a holistic view of how all the systems are constructed. In some senses you and I use the services of every Systems Researcher, as they are usually the architects of the systems we use (and run code on). Take content delivery over the network as an example. Imagine the amount of applications, videos, music which is viewed over the internet every second; surely entities like cnn.com would have had headaches distributing this content. But thanks to some systems researchers, the founders of Akamai delivered (and published) a solution, which probably now lets you keep in touch with news without hiccups.

tinkering Google's MapReduce (Current Research)

Let me give you another example where Systems Research is used everyday. Take the Google search engine for example. It takes close to 0.1 billion queries each day, simply because it enables you to find the anthology of information on the web, faster than our "dog" can search our PCs. Since building and maintaining such data is not trivial, Google published their workings not long ago with the name of MapReduce. Currently we are improving the model so you might like to see what our work is all about.

Advanced Topics in Computer Systems

The course taught by Dr. Umar Saif, although wasn't a hands-on course towards some system implementation, but I still, none-the-less, enjoyed it cause it forced me to think from a System designer's perspective. Interesting stories from Dr. Umar about the people working at the cutting-edge of CS, coupled with a descent set of seminal Systems papers, this course topped the list of what taught me to be a system's researcher :). Papers I thoroughly enjoyed:

  1. TAG: a Tiny Aggregation Service of Ad-Hoc Sensor Networks | S. Madden, M. J. Franklin, J. Hellerstein, W. Hong | Running distributed declarative queries over wireless sensor nodes | Presentation
  2. DDoS Defense By Offense | M. Walfish, M.i Vutukuru, H. Balakrishnan, D. Karger, S. Shenker | Proposal of speak-up, which protects the server-side by exploiting the variance in attitudes of systems with truly legitimate requests and those launching Distributed Denial of Service attacks | Presentation
  3. Some Foundational Problems in Inter-domain Routing | N. Feamster, H. Balakrishnan, J. Rexford | Discussion on flaws in the Inter-domain Routing using Border-Gateway Protocol from a topological, salability and security perspective.
  4. MapReduce: Simplified Data Processing on Large Clusters | J. Dean, S. Ghemawat | A distributed data processing framework used at Google (read 'used extremely successfully') using the map and reduce primitives taken from functional languages | Presentation | See my MapReduce research page
  5. Bigtable: A Distributed Storage System for Structured Data | F. Chang, J. Dean, S. Ghemawat, W. C. Hsieh, D. A. Wallach, M. Burrows, T. Chandra, A. Fikes, R. E. Gruber | High performance distributed data storage model used by Google to store web-indexes, apart from other things | Presentation | Video
  6. Chord: A Scalable Peer-to-Peer Lookup Service for Internet Applications | I. Stoica, R. Morris, D. Karger, M. F. Kaashoek, H. Balakrishnan | One of the original DHT protocols developed at the same time as Berkeley's Scalable CAN | Presentation | Project Page
  7. SplitStream: High Bandwidth Multicast in Cooperative Environments | M. Castro, P. Druschel, A. Kermarrec, A. Nandi, A. Rowstron, A. Singh | Efficient load divisioning in p2p messaging multi-cast systems by moving burden off the interior nodes | Presentation | Project Page
  8. Xen and the Art of Virtualization | P. Barham, B. Dragovic, K. Fraser, S. Hand, T. Harris, A. Ho, R. Neugebauer, I. Pratt, A. Warfield | Provides a Xen Hypervisor to run virtual OSs on a single server. If interested look at Mendel Rosenblums's work too | Presentation | Project Page

By the way, this course has went a long way for me. I eventually ended up tinkering and implementing a large-scale distributed processing system (largely derived from MapReduce). Check out my MapReduce research page

my Systems Links

 
mail: ahmad.humyn@gmail.com | skype: ahmad.humayun | voice: +92 321 4457315