d7008340c8fd9d8e.tex
1: \begin{abstract}
2: We introduce a new family of Fountain codes that are systematic and also have
3: sparse parities.
4: Given an input of $k$ symbols, our codes produce an unbounded number of output
5: symbols, generating each parity independently by linearly combining a
6: logarithmic number of randomly selected input symbols.
7: The construction guarantees that for any $\epsilon>0$ accessing a random subset
8: of $(1+\epsilon)k$ encoded symbols, asymptotically suffices to recover the
9: $k$ input symbols with high probability.
10: 	
11: Our codes have the additional benefit of logarithmic locality: a single lost
12: symbol can be repaired by accessing a subset of $O(\log k)$ of the remaining
13: encoded symbols.
14: This is a desired property for distributed storage  systems where symbols  are
15: spread over a network of storage nodes.
16: Beyond recovery upon loss, local reconstruction provides an efficient
17: alternative for reading symbols that cannot be accessed directly.
18: In our code, a logarithmic number of disjoint local groups is associated
19: with each systematic symbol, allowing \emph{multiple parallel reads}.
20: 
21: Our main mathematical contribution involves analyzing the rank of sparse random matrices with specific structure over finite fields.
22: We rely on establishing that a new family of sparse random bipartite graphs have perfect matchings with high probability. 
23: \end{abstract}
24: