Sergey Gorinsky

The utility of feedback in layered multicast congestion control

Layered multicast is a common approach for dissemination of audio and video in heterogeneous network environments. Layered multicast schemes can be classified into two categories - feedback-based and feedback-free - depending on whether or not the scheme delivers feedback to the sender of the multicast session. Advocates of feedback-based schemes claim that feedback is necessary to match the heterogeneous receiver capabilities efficiently. Supporters of feedback-free schemes believe that feedback introduces significant complexity and that a moderate amount of additional layers can balance any benefit the feedback provides. Surprisingly, there has been no systematic evaluation of these claims. This paper provides a quantitative comparison of feedback-based and feedback-free layered multicast schemes with respect to aligning the provided service to the capabilities of heterogeneous receivers. We discover realistic scenarios when feedback-free schemes require a very large number of additional layers to match the performance of feedback-based schemes. Our studies also demonstrate that a light-weight feedback-based scheme can offer substantial improvement in performance over feedback-free schemes and can closely approximate the efficiency achieved by the optimal feedback-based scheme.

Robustness to inflated subscription in multicast congestion control

Group subscription is a useful mechanism for multicastcongestion control: RLM, RLC, FLID-DL, and WEBRC form a promisingline of multi-group protocols where receivers provide no feedbackto the sender but control congestion via group membershipregulation. Unfortunately, the group subscription mechanism alsooffers receivers an opportunity to elicit self-beneficial bandwidthallocations. In particular, a misbehaving receiver canignoreguidelines for group subscription and choose an unfairly highsubscription level in a multi-group multicast session. This poses aserious threat to fairness of bandwidth allocation. In this paper,we present the first solution for the problem of inflatedsubscription. Our design guards access to multicast groups withdynamic keys and consists of two independent components: DELTA(Distribution of ELigibility To Access) -- a novel method forin-band distribution of group keys to receivers that are eligibleto access the groups according to the congestion control protocol,and SIGMA (Secure Internet Group Management Architecture) -- ageneric architecture for key-based group access at edgerouters.

Robustness of multicast congestion control to inflated subscription

The Internet has grown from a small testbed shared by a close-knit community of researchers to a global commercial network with a huge number of users. The change in the scale requires revisiting original assumptions in the Internet design and checking whether they match the reality of today. One such assumption is trust. Conventional protocols for network bandwidth allocation assume that all communicating parties follow guidelines for fair bandwidth sharing. However, a selfish receiver has incentives to acquire data at an unfairly high rate. Furthermore, open-source operating systems create ample opportunities for receiver misbehavior. For example, Savage et al show that a misbehaving TCP receiver can increase its reliable throughput substantially at the expense of competing traffic [6]. Thus, network research faces a new important challenge of robust bandwidth allocation in the presence of distrusted receivers [2]. Multicast is a service for scalable dissemination of data to a group of receivers: a receiver subscribes to the group by submitting the group address to the local edge router, and the network forwards the data to subscribed receivers. Scalable protocols for multicast congestion control require additional mechanisms, e.g., for feedback aggregation or suppression. To address heterogeneity of receivers, multicast protocols compose a session from multiple groups and provide each receiver with guidelines for choosing a fair subscription level in the session [5]. Resolving the multicast-specific challenges of scalability and heterogeneity opens new opportunities for congestion control misbehavior of receivers. In this work, we focus on an attack of inflated subscription where a receiver ignores subscription guidelines to acquire