No abstract

No abstract

The most common broadband access technology today is the (Asymmetric) Digital Subscriber Line (A)DSL. As the name implies, ADSL allocates the bandwidth of upstream and downstream asymmetrically, significantly favoring the downstream direction. Popular applications that serve content over these networks in the upstream direction, such as peer-topeer (P2P) applications, efficiently utilize the comparably small uplink capacity but potentially disturb inelastic traffic (e.g. voice or games) with an intolerable delay penalty. Effectively, this can render these applications useless.In order to solve this problem, a new delay-based congestion control for this type of P2P background traffic has been proposed. It attempts to utilize the full upstream capacity in the absence of other traffic, otherwise yields to that traffic. This paper describes an implementation and analysis of this new congestion control algorithm in a realistic DSL laboratory setup using a production Multiservice Access Node (MSAN). Using a variety of DSL modems and settings, this worse-than-best-effort transport protocol is further compared to alternative means solving the same problem.

A vast amount of applications and mechanisms recently developed for mobile ad-hoc networks could greatly benefit from the utilization of network status information. That includes, but is not limited to the detection of network partitioning. Network partitioning is a form of network failure. A single connected network topology breaks apart into two or more network topologies separated from each other. Nodes within each partition are still able to communicate with each other but nodes in other partitions are unreachable. This paper proposes two different partition detection mechanisms, one using a centralized approach, the other one utilizing the advantages of a distributed mechanism. The simulations show that both approaches detect partitioning reliably, with both having unique advantages.

No abstract