A Concurrent Partial Snapshot Algorithm for Large-Scale and Dynamic Distributed Systems

Abstract

SUMMARYCheckpoint-rollback recovery, which is a universal method for restoring distributed systems after faults, requires a sophisticated snapshot algorithm especially if the systems are large-scale, since repeatedly taking global snapshots of the whole system requires unacceptable communication cost. As a sophisticated snapshot algorithm, a partial snapshot algorithm has been introduced that takes a snapshot of a subsystem consisting only of the nodes that are communication-related to the initiator instead of a global snapshot of the whole system. In this paper, we modify the previous partial snapshot algorithm to create a new one that can take a partial snapshot more efficiently, especially when multiple nodes concurrently initiate the algorithm. Experiments show that the proposed algorithm greatly reduces the amount of communication needed for taking partial snapshots.

A cooperative partial snapshot algorithm for checkpoint‐rollback recovery of large‐scale and dynamic distributed systems and experimental evaluations

“…In this section, we evaluate the performance of the proposed algorithm with the CSS algorithm. 10,20 The CSS algorithm is a representative of partial snapshot algorithms, as described in Section 2, and the two algorithms have the same properties: (1) The algorithms do not suspend an application execution on a distributed system while taking a snapshot, (2) the algorithms take partial snapshots (not snapshots of the entire system), ( 3) the algorithms can take multiple snapshots concurrently, and (4) the algorithms can handle dynamic network topology changes. In addition, both algorithms are based on the SSS algorithm.…”

“…In contrast, SSS algorithm allows execution of any applications while a snapshot is taken, with some elaborate operations based on the communication-relation. Kim et al, proposed a new partial snapshot algorithm, named Concurrent Sub-Snapshot (CSS) algorithm [11,21], based on SSS algorithm. They called the problematic situation caused by the overlap of the subsystems a collision and presented an algorithm that can resolve collisions by combining colliding SSS algorithm instances.…”

“…Before showing the simulation results, we briefly explain CSS algorithm. For details, please refer the original paper [21].…”

A Distributed Snapshot Protocol for Efficient Artificial Intelligence Computation in Cloud Computing Environments

“…For the practical implementation of the snapshot protocol, the system model must consider failures and asynchronous communication [27]. In [28], a partial snapshot algorithm for a subsystem, where multiple nodes concurrently initiate the snapshot algorithm, is proposed. In Snapify [29], a snapshot algorithm for offload applications on Xeon Phi manycore processors is proposed.…”

“…Algorithm 1 shows the pseudocode of the proposed distributed snapshot algorithm for the active thread. Before starting a round, node i checks whether a consistent global state is collected for failedRound (lines [16][17][18][19][20][21][22][23][24][25][26][27][28]. If the stateNodes data structure satisfies the conditions of the GS, node i saves the stateNodes data structure to latestSnapshot and builds the stateChannel data structure (lines 17-21).…”