Batch Codes for Asynchronous Recovery of Data

Abstract

We propose a new model of asynchronous batch codes that allow for parallel recovery of information symbols from a coded database in an asynchronous manner, i.e. when queries arrive at random times and they take varying time to process. We show that the graph-based batch codes studied by et al. are asynchronous. Further, we demonstrate that hypergraphs of Berge girth larger or equal to 4, respectively larger or equal to 3, yield graph-based asynchronous batch codes, respectively private information retrieval (PIR) codes. We prove the hypergraph-theoretic proposition that the maximum number of hyperedges in a hypergraph of a fixed Berge girth equals the quantity in a certain generalization of the hypergraph-theoretic (6,3)-problem, first posed by Brown, Erdős and Sós. We then apply the constructions and bounds by Erdős, Frankl and Rödl about this generalization of the (6,3)-problem, known as the (3$\varrho$-3,$\varrho$)-problem, to obtain batch code constructions and bounds on the redundancy of the graph-based asynchronous batch and PIR codes. We derive bounds on the optimal redundancy of several families of asynchronous batch codes with the query size $t=2$. In particular, we show that the optimal redundancy $\rho (k)$ of graph-based asynchronous batch codes of dimension $k$ for $t=2$ is $2\sqrt{k}$. Moreover, for graph-based asynchronous batch codes with $t\ge 3$, $\rho (k)=O\left(k^{1/(2-\epsilon )}\right)$ for any small $\epsilon >0$.

18 pages