A survey of rollback-recovery protocols in message-passing systems

By E. Elnozahy, Lorenzo Alvisi, Yi-Min Wang, David B. Johnson. CSUR, 2002

1,952 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Hardware and Architecture

View Paper: https://www.semanticscholar.org/paper/6f9633e70f7027a982fbe6d966b8ca1946212399

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

In Hardware and Architecture

Citations in the corpus

MPICH-V: Toward a Scalable Fault Tolerant MPI for Volatile Nodes (2002, neutral): “As discussed in [8, 21], many techniques of fault tolerance have been proposed in previous works, lying in various level of the software stacks : one solution is based on global coherent checkpoints, the MPI application being restarted to the last coherent checkpoint in case of faults (even in case…”
Designing and Implementing Malicious Hardware (2008, methodology): “We could reverse state changes by us-ing an undo log [7], but restoring system-level check-points would cause timing perturbations.”
Vigilante: end-to-end containment of internet worms (2005, background): “There is a large body of work in this area [14, 13] that we could leverage.”
SafetyNet: improving the availability of shared memory multiprocessors with global checkpoint/recovery (2002, background): “Coordinated system-wide checkpointing avoids both cascading rollbacks [15] and an output commit problem [16] for inter-node communication.”
Fault-tolerance in the Borealis distributed stream processing system (2005, methodology): “We argue that traditional approaches to record reconciliation [27, 42] are illsuited for streaming systems, and adapt two approaches similar to known checkpoint/redo and undo/redo schemes [18, 23, 22, 29, 39] to allow SPEs to reconcile their states.”
The Stratosphere platform for big data analytics (2014, background): “The propagation of restarts is similar to rollback propagation in uncoordinated checkpoint environments [24,66] and has been adopted, in one form or another, by many data flow engines that execute analytical queries: Hadoop [5] (where the data flow is the simple fixed MapShuffle-Reduce pipeline), Dr”
PLFS: a checkpoint filesystem for parallel applications (2009, background): “A broad evaluation of rollback-recovery protocols and checkpointing and a discussion of their future at the petascale can be found in [10, 11].”
ReVirt: enabling intrusion analysis through virtual-machine logging and replay (2002, methodology): “This technique is used commonly in message-logging recovery protocols [ Elnozahy02 ] and can be viewed as expanding the domain of the replay system to include other computers.”
Addressing failures in exascale computing (2014, background): “Many variants of message-logging protocols have been developed (Elnozahy et al., 2002).”
A Survey on Distributed Machine Learning (2019, background): “The second reason is the resilience against failures because, when a single processor fails within an HPC application, the system can still continue operating by initiating a partial recovery (e.g., based on communication-driven checkpointing [46] or partial re-computation [169]).”
Evaluating the viability of process replication reliability for exascale systems (2011, background): “Instead, HPC applications have relied primarily on rollback recovery techniques [13], particularly coordinated checkpoint/restart, where application state is periodically written to stable storage (checkpointed), and when failures occur, this state is used to recover the application to a previously ”
Toward Exascale Resilience (2009, background): “Concerning fault tolerant protocols, the progress in the past 10 years is substantial (Elnozahy et al. 2002; Bouteiller et al. 2008).”
Understanding failures in petascale computers (2007, background): “Storage failures Our interest in large-scale cluster failure originates in the key role of high bandwidth storage in checkpoint/restart strategies for application fault tolerance [11].”
Execution replay of multiprocessor virtual machines (2008, background): “00 of a system just before a crash to be recovered without the need for frequent checkpoints[7, 4, 11].”
High-availability algorithms for distributed stream processing (2005, background): “The primary accumulates such information, called determinants [9]1, in a log message.”
Rx: treating bugs as allergies---a safe method to survive software failures (2005, methodology): “The most straightforward method in this category is to checkpoint, rollback upon failures, and then re-execute either on the same machine [24, 43] or on a different machine designated as the”