An ideal datacenter network should provide several properties, including low median and tail latency, high utilization (throughput), fair allocation of network resources between users or applications, deadline-aware scheduling, and congestion (loss) avoidance. Current datacenter networks inherit the principles that went into the design of the Internet, where packet transmission and path selection decisions are distributed among the endpoints and routers. Instead, we propose that each sender should delegate control---to a centralized arbiter---of when each packet should be transmitted and what path it should follow.

Spinal codes are a recently proposed capacity-achieving rateless code. While hardware encoding of spinal codes is straightforward, the design of an efficient, high-speed hardware decoder poses significant challenges. We present the first such decoder. By relaxing data dependencies inherent in the classic M-algorithm decoder, we obtain area and throughput competitive with 3GPP turbo codes as well as greatly reduced latency and complexity. The enabling architectural feature is a novel alpha-beta incremental approximate selection algorithm. We also present a method for obtaining hints which anticipate successful or failed decoding, permitting early termination and/or feedback-driven adaptation of the decoding parameters.

Recently, rateless codes have introduced a promising approach to obtaining wireless throughput higher than what is achieved by fixed-rate codes, especially over time-varying channels. Rateless codes like Raptor, Strider, and spinal codes naturally process all the information available at the receiver corresponding to a packet, whether from one or many frame transmissions. However, a profitable deployment of rateless codes in a wireless network requires a link-layer protocol to coordinate between sender and receiver. This protocol needs to determine how much coded data should be sent before the sender pauses for feedback from the receiver. Without such feedback, an open-loop sender would not know when the packet has been decoded, but sending this feedback is not free and consumes a significant fraction of the packet transmission time. This paper develops RateMore, a protocol that learns the probability distribution of the number of symbols required to decode a packet (the decoding CDF), and uses the learned distribution in a dynamic programming strategy to produce an optimal transmission schedule. Our experiments show that RateMore reduces overhead by between 2.6x and 3.9x compared to 802.11-style ARQ and between 2.8x and 5.4x compared to 3GPP-style "Try-after-n" HARQ.

Spinal codes are a new class of rateless codes that enable wireless networks to cope with time-varying channel conditions in a natural way, without requiring any explicit bit rate selection. The key idea in the code is the sequential application of a pseudo-random hash function to the message bits to produce a sequence of coded symbols for transmission. This encoding ensures that two input messages that differ in even one bit lead to very different coded sequences after the point at which they differ, providing good resilience to noise and bit errors. To decode spinal codes, this paper develops an approximate maximum-likelihood decoder, called the bubble decoder, which runs in time polynomial in the message size and achieves the Shannon capacity over both additive white Gaussian noise (AWGN) and binary symmetric channel (BSC) models. Experimental results obtained from a software implementation of a linear-time decoder show that spinal codes achieve higher throughput than fixed-rate LDPC codes, rateless Raptor codes, and the layered rateless coding approach of Strider, across a range of channel conditions and message sizes. An early hardware prototype that can decode at 10 Mbits/s in FPGA demonstrates that spinal codes are a practical construction.

A fundamental problem in wireless networks is to develop communication protocols that achieve high throughput in the face of noise, interference, and fading, all of which vary with time. An ideal solution is a rateless wireless system, in which the sender encodes data without any explicit estimation or adaptation, implicitly adapting to the level of noise or interference. In this paper, we present a novel rateless code, the spinal code, which uses a hash function over the message bits to produce pseudo-random bits that in turn can be mapped directly to a dense constellation for transmission. Results from theoretical analysis and simulations show that spinal codes essentially achieve Shannon capacity, and out-perform best-known fixed rate block codes.