BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Load balancing by network curvature control - Professor Edmond Jon
 ckheere (Ming Hsieh Department of Electrical Engineering and Department of
  Mathematics\, University of Southern California)
DTSTART:20090626T130000Z
DTEND:20090626T140000Z
UID:TALK18813@talks.cam.ac.uk
CONTACT:Dr Guy-Bart Stan
DESCRIPTION:It is argued that traffic congestion in computer network is a 
 sequel of the combination of greedy routing and negative curvature. Negati
 ve curvature here is to be interpreted in the sense of Gromov\, which roug
 hly means that the Internet can be approximated by a Riemannian manifold o
 f negative curvature. We will propose a general conjecture that the point 
 of heaviest congestion in a negatively curved network is the center of mas
 s of the network\, defined as a point relative to which the inertia of the
  network is minimum. Next\, if negative curvature implies congestion\, it 
 turns out that most elementary techniques will be ineffective unless they 
 manage to go around the fundamental negative curvature limitation. The pro
 posed curvature based load balancing consists in running the so-called Yam
 abe flow algorithm—instrumental in the proof of the Poincare conjecture
 —to assign link weights so that the resulting network has uniform positi
 ve curvature\, assuming that an Euler characteristic obstruction vanishes.
  Then doing the routing on the modified network with controlled curvature 
 and mapping the routing back to the original network provides nearly unifo
 rm traffic load. 
LOCATION:Cambridge University Engineering Department\, Lecture Room 5
END:VEVENT
END:VCALENDAR