Unformatted text preview: b) What is the key, title, and authors of the paper? Key: [Byers98a] Title: A Digital Fountain Approach to Reliable Distribution of Bulk Data Authors: John W. Byers, Michael Luby, Michael Mitzenmacher, Ashutosh Rege c) what are the key ideas or claims of the paper (state the idea and explain it briefly in 1-2 sentences)? This paper proposes to use digital fountain, a fully scalable protocol based on fast erasure codes to distribute bulk of data to a large number of consumers at times of their choosing. The idea is to ensure reliability without any feedback even in the face of high loss rate, variable link latencies and diverse networking scenarios. d) what is novel about the paper (state the idea and why it is novel, 1 sentence; note that this may be different than your answer to (a))? The authors have proposed a better (more ideal) implementation of digital fountain using a new class of codes (tornado codes) which are much faster, efficient and tolerant to many factors like file sizes, number of consumers and link latencies. e) what methodology (or methodologies) does the paper use (1-2 sentences)? The authors compared tornado codes to Reed-Solomon and Interleaved codes in terms of decoding inefficiencies, encoding/decoding times, and varying file sizes for different packet loss rates. They have compared different coding schemes for implementing digital fountain approach using theoretical examples, simulations and have provided an experimental implementation of reliable distribution protocol using IP Multicast. f) what is the best way the paper could best be improved (1-3 sentences)? (Ideally, improvements would be new work that extend, generalize, or fix the prior paperfuture work. Alternatively, the improvements could a specific better way to present the material.) Very little coverage is provided of the effects on simulations and results when different parameter sets for tornado codes are chosen. Theoretical introduction of tornado codes needs to be a little more descriptive and could yield better understanding if mixed with the example in the next heading. g) Why do you consider this paper important or not, and why (1-2 sentences)? This paper is important because the need to distribute data to a large consumer base has grown by leaps and bounds in the past recent years and efficient use of communication links is vital for enabling networking infrastructure to handle new broadcast-based applications demands....
View Full Document
This note was uploaded on 08/05/2009 for the course CSCI 551 taught by Professor Billcheng during the Spring '08 term at USC.
- Spring '08