Extending R2P2 with Congestion Control and Request-Level Scheduling

Distributed low-latency datacenter applications must meet strict service-level objectives in the order of microseconds. R2P2 is a novel datacenter transport protocol designed for such services. This work extends R2P2 with congestion control and request-level scheduling. It provides a configurable upper bound to in-network queuing using the idleness of the ToR-to-host link as the congestion signal. It also reduces the mean and tail RPC latency by minimizing head-of-line blocking in the network through shortest remaining processing time scheduling. The receiver driven scheduling mechanism is general and can support a variety of scheduling policies. The extension improves the performance of R2P2 under network load by at least 23% and up to two orders of magnitude. Finally, it enables R2P2 to perform similarly to a theoretically near-optimal, clean-slate transport protocol design.

Global Fairness of additive-increase and multiplicative-decrease with heterogeneous round-trip times

Catherine Boutremans, Jean-Yves Le Boudec, Milan Vojnovic

Consider a network with an arbitrary topology and arbitrary communication delays, in which congestion control is based on additive-increase and multiplicative-decrease. We show that the source rates tend to be distributed in order to maximize an objective function called

F_A^h

(``

F_A^h

fairness). We derive this result under the assumption of rate proportional negative feedback and for the regime of rare negative feedback. This applies to TCP in moderately loaded networks, and to those TCP implementations that are designed to interpret multiple packet losses within one RTT as a single congestion indication and do not rely on re-transmission timeout. This result provides some insight into the distribution of rates, and hence of packet loss ratios, which can be expected in a given network with a number of competing TCP or TCP-friendly sources. We validate our findings by analyzing the parking lot scenario, and comparing with previous results \cite{floyd-91-b,mathis-97-a}, and an extensive numerical simulation with realistic parameter settings. We apply

F_A^h

fairness to gain a more accurate understanding of the bias of TCP against long round trip times.

1999

Global fairness of additive-increase and multiplicative-decrease with heterogeneous round-trip times

Catherine Boutremans, Jean-Yves Le Boudec, Milan Vojnovic

Consider a network with an arbitrary topology and arbitrary communication delays, in which congestion control is based on additive--increase and multiplicative--decrease. We show that the source rates tend to be distributed in order to maximize an objective function called

F_A^h

(``

F_A^h

fairness`). We derive this result under the assumption of rate proportional negative feedback and for the regime of rare negative feedback. This applies to TCP in moderately loaded networks, and to those TCP implementations that are designed to interpret multiple packet losses within one RTT as a single congestion indication and do not rely on re-transmission timeout. This result provides some insight into the distribution of rates, and hence of packet loss ratios, which can be expected in a given network with a number of competing TCP or TCP-friendly sources. We validate our findings by analyzing the parking lot scenario, and comparing with previous results \cite{floyd-91-b,mathis-97-a}, and an extensive numerical simulation with realistic parameter settings. We apply

F_A^h

fairness to gain a more accurate understanding of the bias of TCP against long round trip times.

2000

Network Resource Allocation for Competing Multiple Description Transmissions

Suhas Diggavi

Providing real-time multimedia services over a best-effort network is challenging due to the stringent delay requirements in the presence of complex network dynamics. Multiple description (MD) coding is one approach to transmit the media over diverse (multiple) paths to reduce the detrimental effects caused by path failures or delay. The novelty of this work is to investigate the resource allocation in a network, where there are several competing MD coded streams. This is done by considering a framework that chooses the operating points for asymmetric MD coding to maximize total quality of the users, while these streams are sent over multiple routing paths. We study the joint optimization of multimedia (source) coding and congestion control in wired networks. These ideas are extended to joint source coding and channel coding in wireless networks. In both situations, we propose distributed algorithms for optimal resource allocation. In the presence of path loss and competing users, the service quality to any particular MD stream could be uncertain. In such circumstances it might be tempting to expect that we need greater redundancy in the MD streams to protect against such failures. However, one surprising aspect of our study reveals that for large number of users who compete for the same resources, the overall system could benefit through opportunistic (hierarchical) strategies. In general networks, our studies indicate that the user composition varies from conservative to opportunistic operating points, depending on the number of users and their network vantage points.

2008

Global fairness of additive-increase and multiplicative-decrease with heterogeneous round-trip times

Catherine Boutremans, Jean-Yves Le Boudec, Milan Vojnovic

F_A^h

(``

F_A^h

F_A^h

fairness to gain a more accurate understanding of the bias of TCP against long round trip times.

2000

Global Fairness of additive-increase and multiplicative-decrease with heterogeneous round-trip times

Catherine Boutremans, Jean-Yves Le Boudec, Milan Vojnovic

F_A^h

(``

F_A^h

F_A^h

fairness to gain a more accurate understanding of the bias of TCP against long round trip times.

1999