Summary of Papers covered:
The first paper's [1] objective is to design overlay services for making informed application specific routing decisions. The authors of this paper reason that routing decisions made to different types of network services like file sharing, content distribution networks, object location services cannot make direct use of information available at the underlying internet directly. This limits the types of routing decisions that the network services can make, given that for example, not all BGP routers expose their route tables for network services to make use of, either for security or business reasons. Further, use traceroute or ping to determine the network topology, which results in un-necessary traffic. Hence, the authors propose a layered architecture, wherein a topology probing kernel collects network information and provides it to a library of routing services, that applies intelligent algorithms to conclude network parameters. These parameters are then provided in a generic fashion to network services, which can make intelligent routing decisions. The drawback of this proposal is that it is too optimistic that routers and network devices will implement this type of a architecture. The second paper [2] discusses server selection and overlay construction based on assumption that applications do not require exact topological information, but need sufficient hints on relative position of internet hosts; by tying overlay construction with underlying internet topology. The primary measure being the latency. This model requires dedicated hosts throughout the internet to calculate latencies to dedicated hosts and perform binning based on measured latencies. The last paper [3] discusses set of algorithms that can be employed to improve routing decisions. The algorithms make use of distance as a cost factor and degree relaxation techniques to realign routes. Overall, in my opinion, routing algorithms have better chance of acceptance, given that they do not require routers to expose information or need dedicated set of hardware throughout the internet for determining latencies.
Links to papers covered
[1] A routing underlay for overlay networks by Akihiro Nakao, Larry Peterson and Andy Bavier; Princeton University
[2] Topologically aware overlay construction and server selection by Sylvia Ratnasamy, Mark Handley, Richard Karp, Scott Shenker
[3] Routing Overlay in Multicast networks by Sherlia Shi, Jonathan Turner; Washington University in St. Louis
Presentation given in class on 2nd March:
cse525_PG20_ChetanHiremath.ppt
Links to other papers and web links:
None