Switch Codes: Codes for Fully Parallel Reconstruction: Codes for Fully Parallel Reconstruction

Zhiying Wang, Han Mao Kiah, Yuval Cassuto, Jehoshua Bruck

Research output: Contribution to journalArticlepeer-review

Abstract

Network switches and routers scale in rate by distributing the packet read/write operations across multiple memory banks. Rate scaling is achieved so long as sufficiently many packets can be written and read in parallel. However, due to the non-determinism of the read process, parallel pending read requests may contend on memory banks, and thus significantly lower the switching rate. In this paper, we provide a constructive study of codes that guarantee fully parallel data reconstruction without contention. We call these codes 'switch codes,' and construct three optimal switch-code families with different parameters. All the constructions use only simple XOR-based encoding and decoding operations, an important advantage when operated in ultra-high speeds. Switch codes achieve their good performance by spanning simultaneous disjoint local-decoding sets for all their information symbols. Switch codes may be regarded as an extreme version of the previously studied batch codes, where the switch version requires parallel reconstruction of all the information symbols.

Original languageEnglish
Article number7843662
Pages (from-to)2061-2075
Number of pages15
JournalIEEE Transactions on Information Theory
Volume63
Issue number4
DOIs
StatePublished - Apr 2017

Keywords

  • Distributed-storage codes
  • batch codes
  • combinatorial designs
  • network switches

All Science Journal Classification (ASJC) codes

  • Information Systems
  • Computer Science Applications
  • Library and Information Sciences

Fingerprint

Dive into the research topics of 'Switch Codes: Codes for Fully Parallel Reconstruction: Codes for Fully Parallel Reconstruction'. Together they form a unique fingerprint.

Cite this