arXiv++ Combinatorics

Batch codes, first introduced by Ishai, Kushilevitz, Ostrovsky, and Sahai, mimic a distributed storage of a set of $n$ data items on $m$ servers, in such a way that any batch of $k$ data items can be retrieved by reading at most some $t$ symbols from each server. Combinatorial batch codes, are replication-based batch codes in which each server stores a subset of the data items. In this paper, we propose a generalization of combinatorial batch codes, called multiset combinatorial batch codes (MCBC), in which $n$ data items are stored in $m$ servers, such that any multiset request of $k$ items, where any item is requested at most $r$ times, can be retrieved by reading at most $t$ items from each server. The setup of this new family of codes is motivated by recent work on codes which enable high availability and parallel reads in distributed storage systems. The main problem under this paradigm is to minimize the number of items stored in the servers, given the values of $n,m,k,r,t$, which is denoted by $N(n,k,m,t;r)$. We first give a necessary and sufficient condition for the existence of MCBCs. Then, we present several bounds on $N(n,k,m,t;r)$ and constructions of MCBCs. In particular, we determine the value of $N(n,k,m,1;r)$ for any $n\geq \left\lfloor\frac{k-1}{r}\right\rfloor{m\choose k-1}-(m-k+1)A(m,4,k-2)$, where $A(m,4,k-2)$ is the maximum size of a binary constant weight code of length $m$, distance four and weight $k-2$. We also determine the exact value of $N(n,k,m,1;r)$ when $r\in\{k,k-1\}$ or $k=m$.

Batch codes, introduced by Ishai, Kushilevitz, Ostrovsky and Sahai, represent the distributed storage of an $n$-element data set on $m$ servers in such a way that any batch of $k$ data items can be retrieved by reading at most one (or more generally, $t$) items from each server, while keeping the total storage over $m$ servers equal to $N$. This paper considers a class of batch codes (for $t=1$), called combinatorial batch codes (CBC), where each server stores a subset of a database. A CBC is called optimal if the total storage $N$ is minimal for given $n,m$, and $k$. A $c$-uniform CBC is a combinatorial batch code where each item is stored in exactly $c$ servers. A $c$-uniform CBC is called optimal if its parameter $n$ has maximum value for given $m$ and $k$. Optimal $c$-uniform CBCs have been known only for $c\in \{2,k-1,k-2\}$. In this paper we present new constructions of optimal CBCs in both the uniform and general settings, for values of the parameters where tight bounds have not been established previously. In the uniform setting, we provide constructions of two new families of optimal uniform codes with $c\sim \sqrt{k}$. Our constructions are based on affine planes and transversal designs.