Su Wen

Lightweight network support for scalable end-to-end services

Some end-to-end network services benefit greatly from network support in terms of utility and scalability. However, when such support is provided through service-specific mechanisms, the proliferation of one-off solutions tend to decrease the robustness of the network over time. Programmable routers, on the other hand, offer generic support for a variety of end-to-end services, but face a different set of challenges with respect to performance, scalability, security, and robustness. Ideally, router-based support for end-to-end services should exhibit the kind of generality, simplicity, scalability, and performance that made the Internet Protocol (IP) so successful. In this paper we present a router-based building block called ephemeral state processing (ESP), which is designed to have IP-like characteristics. ESP allows packets to create and manipulate small amounts of temporary state at routers via short, predefined computations. We discuss the issues involved in the design of such a service and describe three broad classes of problems for which ESP enables robust solutions. We also present performance measurements from a network-processor-based implementation.

A simple loss differentiation approach to layered multicast

Layered multicast is a promising technique for broadcasting adaptive-quality TV video to heterogeneous receivers. While several-layered multicast approaches have been proposed, prior work has identified several problems including significant and persistent instability in video quality, arbitrary unfairness with other sessions, low access link utilization due to conservative bandwidth allocation, and problems with receiver synchronization. In this paper we propose a new layered multicast scheme, where we exploit a simple, coarse-grained, two-tier loss differentiation architecture to achieve stable and fair bandwidth allocation for viewers. Despite the simplicity of our loss differentiation model, we show that it achieves most of the benefits of complex and costly priority dropping schemes. In addition, our protocol is receiver-driven and thus retains the incentives to limit bandwidth usage that are not present in existing priority dropping schemes.

Building multicast services from unicast forwarding and ephemeral state

We present an approach to building multicast services at the network layer using unicast forwarding and two additional building blocks: ephemeral state probes, i.e. extremely lightweight distributed computations based on a time-bounded associative memory; and the ability to inject or enable packet processing functions that modify router behavior in a very limited way. In our approach, senders and receivers use ephemeral state probes to determine where to inject functionality. A special function that duplicates packets matching a particular pattern and forwards them to a specific destination is then instantiated at the desired network location. Our approach eliminates the need for sophisticated multicast routing protocols and gives the end-systems control over the multicast service, allowing the application to tailor the service to its needs. At the same time, our approach creates efficient forwarding paths by using ephemeral state probes to determine (only) the relevant aspects of the network and group topology. We present two multicast implementations: one builds a multicast tree with centralized control, another provides the traditional IP multicast abstraction. Both implementations can be done in a simple and scalable manner with minimal added functionality in the routers beyond unicast forwarding.

Concast: design and implementation of a new network service

This paper introduces concast, a new network service. Concast is the inverse of multicast: multiple sources send messages toward the same destination, which results in a single message being delivered to the destination. The received message appears to come from the concast group rather than any particular receiver. Different forms of concast service can be defined by varying the mapping from the set of sent messages to the received message. The service is useful for preventing implosion and reducing bandwidth consumption in cases where many senders transmit to the same receiver-for example in aggregating (or suppressing) positive (or negative) acknowledgements. We define the semantics of a simple concast service that is the inverse of multicast, as well as a more general custom concast, which allows users to define certain aspects of the services semantics. We describe how to implement the service so that it scales approximately as well as IP multicast. We also present results from a simulation study showing that concast provides significant benefits in a layered-video application.

CALM: congestion-aware layered multicast

This paper demonstrates the flexibility and utility of two lightweight, general-purpose network services (Emphemeral State Processing and LightWeight Processing modules) by showing how end-systems can use these services to obtain timely and accurate information about the location of congested links and their level of congestion. Although accurate congestion information is useful to a wide range of network services, here we illustrate its benefits to layered multicast systems. In particular, we show how these services can be used to overcome well-known problems with layered multicast, including the desire to reduce router state, the need to drop layers quickly (i.e., within one RTT), the problem of coordinating receivers, and the desire to support fine-grained layering without thrashing between layers-even in the face of join experiments.

Integrating concast and multicast communication models

This paper defines a new group communication model called concast communication. Being the counterpart to multicast, concast involves multiple senders transmitting to a single receiver. Concast communication is used in a wide range of applications including collaborative applications, report-in style applications, or just end-to-end acknowledgements in a reliable multicast protocol. This paper explores the issues involved in designing concast communication services. We examine various message combination methods including concatenation, compression, and reduction to reduce the traffic loads imposed on the network and packet implosion at the receiver. Group management operations such as group creation/deletion, joining/leaving, and concast routing are discussed. We also address transmission issues such as reliable delivery, flow control, congestion control, and QoS. We conclude the paper by presenting a concast communication model that we have been developing in the context of TMTP5. The model uses concast communication to implement reliable multicast and it shares concast trees with the multicast group whenever possible to reduce overhead costs.

Leveraging emerging network services to scale multimedia applications

Multicast services have been used to transmit multimedia data to large receiver groups. Only recently have counterpart network services been introduced to provide similar scalability and anonymity in the opposite direction (i.e. messages from a group of senders destined for a common receiver).