Subspace synchronization: A network-coding approach to object reconciliation

Assume that two users possess two different subspaces of an ambient linear space. We show that the problem of synchronization of such vector spaces can be easily solved by an efficient algorithm. By building on this observation, we propose an algorithm for synchronization of two collections of binary files of length n each, stored in the cloud in a distributed manner. By further employing techniques akin to network coding, we propose a more efficient file synchronization algorithm that has communication complexity O(d · n) bits and computational complexity O(k2 · n) operations, where k is the total number of files and d is the number of files that differ. The algorithm successfully reconciles two sets of files in 3 communication rounds with high probability.

[1]  George Varghese,et al.  Biff (Bloom filter) codes: Fast error correction for large data sets , 2012, 2012 IEEE International Symposium on Information Theory Proceedings.

[2]  Frank R. Kschischang,et al.  Coding for Errors and Erasures in Random Network Coding , 2008, IEEE Trans. Inf. Theory.

[3]  Michael T. Goodrich,et al.  Invertible bloom lookup tables , 2011, 2011 49th Annual Allerton Conference on Communication, Control, and Computing (Allerton).

[4]  R. Koetter,et al.  An algebraic approach to network coding , 2001, Proceedings. 2001 IEEE International Symposium on Information Theory (IEEE Cat. No.01CH37252).

[5]  George Varghese,et al.  The complexity of object reconciliation, and open problems related to set difference and coding , 2012, 2012 50th Annual Allerton Conference on Communication, Control, and Computing (Allerton).

[6]  Rudolf Ahlswede,et al.  Network information flow , 2000, IEEE Trans. Inf. Theory.

[7]  What's the difference?: efficient set reconciliation without prior context , 2011, SIGCOMM.

[8]  Yaron Minsky,et al.  Set reconciliation with nearly optimal communication complexity , 2003, IEEE Trans. Inf. Theory.