Consistent Checkpoint Research Articles

An Expert System for Analysis of Consistency Criteria in Checkpointing Algorithms

In a distributed computing environment, it is vital to maintain the states of the processes involved in order to cater to failures that are arbitrary in nature. To reach a consistent state among all the processes, checkpoints are taken locally by each process and are combined together based on uniformity criteria such as consistency, transitlessness, and strong consistency. In this article, first, the necessary and sufficient conditions of consistency criteria are stated and then an expert system, implemented based on these criteria, is presented. The expert system discovers and illustrates consistent, transitless, strongly consistent and globally consistent checkpoints in a given distributed system. Moreover, it offers facilities for evaluating checkpointing algorithms by measuring different quality assessment parameters.

GUARANTEED MUTUALLY CONSISTENT CHECKPOINTING IN DISTRIBUTED COMPUTATIONS

In this paper, we explore the isomorphism between vector time and causality to characterize consistency of a set of checkpoints in a distributed computing. A necessary and sufficient condition, to determine if a set of local checkpoints can form a consistent global checkpoint, is presented and proved using the isomorphic power of vector time and causality. To the best of our knowledge, this is the first attempt to use the isomorphism for this purpose. This condition leads to a simple and straightforward algorithm for a guaranteed mutually consistent global checkpointing. In our approach, a process can take a checkpoint whenever and wherever it wants while other related process may be asked to take an additional checkpoint for ensuring the mutual consistency. We also show how this condition and the resulting algorithm can be used to obtain a maximum and minimum global checkpoints, another important paradigm for distributed applications.

SFT: A consistent checkpointing algorithm with short freezing time

A consistent checkpointing algorithm with short freezing time (SFT) is presented in this paper. It supports fault-tolerance in distributed systems. The algorithm has shorter freezing time, lower overhead, and simplicity of recovery. To make checkpoint time shorter, a special control message (Munblock) is used to ensure that a process can respond the checkpoint event quickly at any given time. Moreover, main memory algorithm is used to improve the concurrency of checkpointing. By using SFT, the freezing time resulted by checkpointing is less than 0.03s. Furthermore, the control message number of SFT is onlyO(n).

Staggered consistent checkpointing

A consistent checkpointing algorithm saves a consistent view of a distributed application's state on stable storage. The traditional consistent checkpointing algorithms require different processes to save their state at about the same time. This causes contention for the stable storage, potentially resulting in large overheads. Staggering the checkpoints taken by various processes can reduce checkpoint overhead. This paper presents a simple approach to arbitrarily stagger the checkpoints. Our approach requires that the processes take consistent logical checkpoints, as compared to consistent physical checkpoints enforced by existing algorithms. Experimental results on nCube-2 are presented.

Supporting Cost-Effective Fault Tolerance in Distributed Message-Passing Applications with File Operations

In this paper we present an approach to reliable distributed computing, which incorporates fault tolerance into applications at low cost, in terms of both run-time performance and programming effort required to construct reliable application software. In our model fault tolerance is based on distributed consistent checkpointing and rollback-recovery integrated with a user-level reliable transmission protocol. By employing novel techniques 8and algorithms, our approach is distinguished from other consistent checkpointing schemes by the following features: first, minimum communication overhead for constructing a consistent distributed checkpoint and catching messages in transit during checkpointing; second, tolerance to message losses due to site failures or unreliable non-FIFO networks; and third, efficient checkpointing and recovery of persistent state, i.e., user files. Based on the model, a software library prototype called Libra has been implemented for supporting fault tolerance in distributed message-passing applications with file operations. The library provides an easy to use programming interface including message-passing and file I/O primitives, which hides the complexity of both fault-tolerant network communications and checkpointing and recovering user files from the application level. Experience with a number of long-running distributed applications shows that Libra can provide fault tolerance in a cost-effective manner.

SFT

SFT algorithm, a consistent checkpointing algorithm with shorter freezing time, is presented in this paper. SFT is able to implement fault-tolerance in distributed systems. The features of the algorithm include shorter freezing time, lower overhead, and simple roll backing. To reduce checkpointing time, a special control message (Munblock) is used to ensure that at any given time a process can respond the checkpoint event quickly. Moreover, a main memory algorithm is used to improve concurrency of checkpointing. By using SFT algorithm, the freezing time resulted by checkpointing is less than 0.03s. The control message number of SFT is only O (n).

An efficient approach for constructing reliable distributed applications

A trend of supporting fault tolerance in distributed computing systems is to incorporate fault tolerance into applications at low cost, in terms of both run-time performance and programming effort required to construct reliable application software. To meet this objective, we have designed and implemented Libra, a software library running on a network of workstations to support reliable distributed computing efficiently. By providing fault tolerance transparency and a simple, easy to use high-level message-passing interface, Libra simplifies the development of reliable distributed applications. Fault tolerance is based on distributed consistent checkpointing and rollback-recovery integrated with a user-level network communication protocol. To address real-life problems, the library tolerates messages losses, and efficiently supports checkpointing and recovery of user files. Performance evaluations show that Libra imposes low run-time overhead, and minimizes communication overhead for taking a consistent distributed checkpoint and catching messages in transit during checkpointing.

A Survey of Distributed Database Checkpointing

Checkpointing a database is a vital technique to reduce the recovery time in the presence of a failure. For distributed databases, checkpointing also provides an efficient way to perform global reconstruction. In this paper, we survey and classify previous approaches for checkpointing a distributed database. Since the need for global reconstruction is infrequent in most distributed databases, a less restrictive and less resource-consuming approach to checkpoint distributed databases in an integrated distributed database system is recommended over a transaction consistent checkpoint approach. For a federated or multidatabase system, any type of global consistent checkpoint is difficult to achieve without violating local autonomy.

Efficient checkpointing procedures for fault tolerant distributed systems

A classical approach for achieving fault tolerance in distributed systems is based on the incorporation of efficient and fault tolerant procedures for checkpointing and recovery in such systems. We propose two checkpointing procedures, which can be initiated by any process in the system or upon failure of one or more component processes. Our procedures return the most recent and consistent checkpoints for the processes initiating the procedure, and do not interfere with the progress of the distributed system application. Furthermore, our procedures guarantee that a consistent checkpoint will be obtained when they terminate. Examples illustrating the application of the procedures are also provided.

Local states in distributed computations

If events produced by processes of a distributed computation are generally supposed to be instantaneous, it is not the case for local states generated by these events. Due to message exchanges and synchronization local states have some duration. This paper defines notions about local states such as weak precedence, strong precedence, weak concurrency and strong concurrency . Moreover a few formulas based on vector clocks, and consequently usable in an operational context, are introduced to decide about relations between local states. These relations and formulas can be used either to debug, test or analyze distributed programs (especially for global properties detection) or to define consistent checkpoints.

Ickp: a consistent checkpointer for multicomputers

There has been much research on checkpointing algorithms for parallel and distributed systems; but surprisingly few implementations for uniprocessors, multiprocessors, and distributed systems, and none at all for multicomputers. We discuss ickp, our consistent checkpointer for the Intel iPSC/860, which is the first general-purpose checkpointer for a multicomputer. It is a checkpointing library that may be invoked asynchronously from the host processor, at a periodic interval, or by a library call. It implements three consistent checkpointing algorithms, two optimizations to reduce checkpoint time and overhead, and recovery.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Checkpointing and Rollback-Recovery for Distributed Systems

We consider the problem of bringing a distributed system to a consistent state after transient failures. We address the two components of this problem by describing a distributed algorithm to create consistent checkpoints, as well as a rollback-recovery algorithm to recover the system to a consistent state. In contrast to previous algorithms, they tolerate failures that occur during their executions. Furthermore, when a process takes a checkpoint, a minimal number of additional processes are forced to take checkpoints. Similarly, when a process rolls back and restarts after a failure, a minimal number of additional processes are forced to roll back with it. Our algorithms require each process to store at most two checkpoints in stable storage. This storage requirement is shown to be minimal under general assumptions.