Download PDFOpen PDF in browser

A Low Overhead Logging Scheme for Fast Recovery in Distributed Shared Memory Systems

EasyChair Preprint 3517

22 pagesDate: May 30, 2020

Abstract

This  paper  presents  an  efficient,  writer-based  logging  scheme  for  recoverable  distributed  shared  memory  systems,  in  which  logging  of  a  data  item  is  performed  by  its  writer  process,  instead  of  every  process  that  accesses  the  item  logging  it.  Since  the  writer  process  maintains  the  log  of  data  items,  volatile  storage  can  be  used  for  logging.  Only  the  readers’  access  information  needs  to  be  logged  into  the  stable  storage  of  the  writer  process  to  tolerate  multiple  failures.  Moreover,  to  reduce  the  frequency  of  stable  logging,  only  the  data  items  accessed  by  multiple  processes  are  logged  with  their  access  information  when  the  items  are  invalidated,  and  also  semantic-based  optimization  in  logging  is  considered.  Compared  with  the  earlier  schemes  in  which  stable  logging  was  performed  whenever  a  new  data  item  was  accessed  or  written  by  a  process,  the  size  of  the  log  and  the  logging  frequency  can  be  significantly  reduced  in  the  proposed  scheme

Keyphrases: Check pointing, fault-tolerant system, message logging, rollback-recovery and log size, shared memory system

BibTeX entry
BibTeX does not have the right entry for preprints. This is a hack for producing the correct reference:
@booklet{EasyChair:3517,
  author    = {Mushtaq Ahmad and Saima Amir and Mehwish Shabbir and Muhammad Nadeem},
  title     = {A  Low  Overhead  Logging  Scheme  for  Fast  Recovery  in  Distributed  Shared  Memory  Systems},
  howpublished = {EasyChair Preprint 3517},
  year      = {EasyChair, 2020}}
Download PDFOpen PDF in browser