Periodic Checkpointing Research Articles

Dynamic Adaptation of Checkpoints and Rescheduling in Grid Computing

is a form distributed computing mainly to virtualilze and utilize geographically distributed idle resources. A grid is a distributed computational and storage environment often composed of heterogeneous autonomously managed subsystems. As a result varying resource availability becomes common place, often resulting in loss and delay of executing jobs. To ensure good performance fault tolerance should be taken into account. Here we address the fault tolerance in terms of resource failure. Commonly utilized techniques to achieve fault tolerance is periodic checkpointing, which periodically saves the jobs state. But an inappropriate checkpointing interval leads to delay in the job execution, and reduces the throughput. Hence in the proposed work, the strategy used to achieve fault tolerance is by dynamically adapting the checkpoints based on current status and history of failure information of the resource, which is maintained in the Information server. The Last failure time and Mean failure time based algorithm dynamically modifies the frequency of checkpoint interval, hence increases the throughput by reducing the unnecessary checkpoint overhead. In case of resource failure, the proposed Fault Index Based Rescheduling (FIBR) algorithm reschedules the job from the failed resource to some other available resource with the least Fault-index value and executes the job from the last saved checkpoint. This ensures the job to be executed within the deadline with increased throughput and helps in making the grid environment trust worthy.

Adaptive Fault Management of Parallel Applications for High-Performance Computing

As the scale of high performance computing (HPC) grows, application fault resilience becomes increasingly important. In this paper, we propose FT-Pro, an adaptive fault management approach that combines the merits of reactive checkpointing and proactive migration. It enables parallel applications to avoid anticipated failures via preventive migration, and in the case of unforeseeable failures, to minimize their impact through selective checkpointing. An adaptation manager is designed for making runtime decision in response to failure prediction. We evaluate FT-Pro through stochastic modeling and case studies with real applications under a wide range of settings. Preliminary results indicate that FT-Pro outperforms periodic checkpointing, in terms of both reducing application completion times and improving resource utilization, by up to 43%.

Analytical study of migration-enhanced fault tolerance for long-running applications in IFR systems

Computer systems with increasing failure rate (IFR) are common in practice. For such systems, the literature indicates that aperiodic checkpointing can provide better performance than periodic checkpointing due to its adaptability to the failure process. However, for long-running applications, aperiodic checkpointing suffers from substantial operational overhead due to frequent checkpointing operations as the application proceeds. To address this problem, in this paper, we propose to incorporate just-in-time process migration in addition to aperiodic checkpointing for applications running in an IFR system. The goal is to reduce application execution time in the presence of failures. In particular, we present an analytical study of this migration-enhanced fault tolerance scheme (denoted as migCP) by deriving application completion time by using migCP and further determining the optimal migration locations. We demonstrate, through analytical modelling and empirical studies, that migCP outperforms aperiodic checkpointing under a variety of system parameters.

실시간 제어 시스템의 결함 극복을 위한 이중화 구조와 체크포인팅 기법의 성능 분석

This paper deals with a performance analysis of real-time control systems, which engages DMR(dual modular redundancy) to detect transient errors and checkpointing technique to tolerate transient errors. Transient errors are caused by transient faults and the most significant type of errors in reliable computer systems. Transient faults are assumed to occur according to a Poisson process and to be detected by a dual modular redundant structure. In addition, an equidistant checkpointing strategy is considered. The probability of the successful task completion in a real-time control system where periodic checkpointing operations are performed during the execution of a real-time control task is derived. Numerical examples show how checkpoiniting scheme influences the probability of task completion. In addition, the result of the analysis is compared with the simulation result.

Exposure to the Sobriety “Checkpoint Strikeforce” Campaign in Maryland: Impact on Driver Perceptions of Vulnerability and Behavior

Objectives: To compare drivers who were: (1) exposed to Maryland's sobriety “Checkpoint Strikeforce” campaign; (2) not exposed but were aware of it; and (3) neither exposed nor aware of it. This six-month campaign involved periodic road-side sobriety checkpoints, in which police attempted to detect and arrest drunk drivers and a series of radio Public Service Announcements (PSAs) that promoted public awareness. Methods: A statewide sample of 1725 drivers were interviewed by telephone to determine their exposure to and awareness of this campaign, as well as their beliefs and behaviors concerning drunk driving and seat belt usage. Results: Relatively few people (9.6%) were exposed in any way to the campaign; however almost 30% were aware of it. Significant increases in perceived vulnerability to arrest for drinking and driving were found among those who were exposed. There were no differences in reported drinking and driving, but seat belt usage in the last 30 days was higher among those who were exposed. There were no differences among those people who had been directly exposed to the campaign (i.e., personally gone through a sobriety checkpoint) and those who were indirectly exposed (i.e., knew a family member or someone else who went through a checkpoint). Conclusions: Exposure to a sobriety checkpoint was associated with increased perceptions of drink-driving risk (i.e., likelihood of getting caught by the police), which was not dependent upon having personally experienced it. Awareness without exposure appears to be insufficient to change perceived risk. The need to increase the reach and frequency of the sobriety “Checkpoint Strikeforce” campaign is indicated.

The interplay of power management and fault recovery in real-time systems

We describe how to exploit the scheduling slack in a real-time system to reduce energy consumption and achieve fault tolerance at the same time. During failure-free operation, a task takes checkpoints to enable recovery from failure. Additionally, the system exploits the slack to conserve energy by reducing the processor speed. If a task fails, it will restart from a saved checkpoint and execute at maximum speed to guarantee that the deadlines are met. We show that the number of checkpoints and their placements interact in subtle ways with the power management policy. We study two checkpoint placement policies for aperiodic tasks and analytically derive the optimal number of checkpoints to conserve energy under each. This optimal number allows the CPU speed to be slowed down to the level that yields minimum energy consumption, while still guaranteeing recoverability of tasks under each checkpointing policy. The results show that traditional periodic checkpointing is not the best policy for the combined purpose of conserving energy and guaranteeing recovery. Instead, better energy savings are possible through a nonuniform distribution of checkpoints that takes into account the energy consumption and reliability factors. Depending on the amount of slack and the checkpointing overhead, energy can be reduced by up to 68 percent under nonuniform checkpointing. We also demonstrate the applicability of these checkpoint placement policies to periodic tasks.

An aggressive approach to failure restoration of PCS mobility databases

In Personal Communications Services (PCS) networks, subscribers' current locations are usually maintained using a two-level hierarchical strategy with two types of databases, specifically, VLRs and HLRs. Location records might be lost due to the malfunction of mobility databases. GSM uses aggressive restoration to restore the HLR data from the VLRs. However, previous work on aggressive restoration is based on periodic checkpointing and might not be feasible in general. This paper proposes a novel aggressive approach, based on aperiodic checkpointing, to back up HLRs and VLRs. The aperiodic checkpointing scheme could use the number of uncheckpointed location records as the threshold value to trigger a backup process. Performance analysis shows that aperiodic checkpointing is more feasible than periodic checkpointing for aggressive restoration.

Increasing processor utilization in hard-real-time systems with checkpoints

Often hard real-time systems require results that are produced on time despite the occurrence of processor failures. This paper considers a distributed system where tasks are periodic and each task occurs in multiple copies which are periodically synchronized in order to handle failures. The problem of preemptively scheduling a set of such tasks is discussed where every occurrence of a task has to be completely executed before the next occurrence of the same task. First, a static scheduling algorithm is proposed which uses periodic checkpoints to tolerate processor failures. Then, the performance of the algorithm is substancially improved employing a mixed strategy which constructs a schedule where high frequency tasks are duplicated, and low frequency tasks are periodically checkpointed. The performance of the solution proposed is evaluated in terms of the minimum achievable processor utilization due to the useful computation of the tasks. Moreover, analytical and simulation studies are used to reveal interesting trade-offs associated with the scheduling algorithm. In particular, if high frequency tasks are less than 70 percent of the total number of tasks then the mixed strategy yields a higher processor utilization than the task duplication scheme.

Recovery approach to the design of stabilizing communication protocols

This paper addresses the problem of designing stabilizing computer communication protocols modelled by communicating finite state machines. A communication protocol is said to be stabilizing if, starting from or reaching any illegal state, the protocol will eventually reach a legal (or consistent) state, and resume its normal execution. To achieve stabilization, the protocol must be able to detect the error as soon as it occurs, and then it must recover from that error and revert back to a legal protocol state. The later issue related to recovery is tackled here, and an efficient procedure for the recovery in communications protocols is described. The recovery procedure does not require periodic checkpointing and, therefore, is less intrusive. It requires less time for rollback and fewer recovery control messages than other procedures. Only a minimal number of processes will roll back, and a minimal number of protocol messages will be retransmitted during recovery. Moreover, our procedure requires minimal stable storage to be used to record contextual information exchanged during the progress of the protocol. Finally, our procedure is compared with an existing recovery procedure, and an illustrative example is provided.

Effects of the checkpoint interval on time and space in time warp

Optimistically synchronized parallel discrete-event simulation is based on the use of communicating sequential processes. Optimistic synchronization means that the processes proceed under the assumption that a synchronized execution schedule is fortuitous. Periodic checkpointing of the state of a process allows the process to roll back to an earlier state when synchronization errors are detected. This article examines the effects of varying the checkpoint interval on the execution time and memory space needed to perform a parallel simulation. The empirical results presented in this article were obtained from the simulation of closed stochastic queuing networks with several different topologies. Various intraprocessor process-scheduling algorithms and both lazy and aggressive cancellation strategies are considered. The empirical results are compared with analytical formulae predicting time-optimal checkpoint intervals. Two modes of operation, throttling and thrashing , have been noted and their effect examined. As the checkpoint interval is increased from one, there is a throttling effect among processes on the same processor, which improves performance. When the checkpoint interval is made too large, there is a thrashing effect caused by interaction between processes on different processors. It is shown that the time-optimal and space-optimal checkpoint intervals are not the same. Furthermore, a checkpoint interval that is too small affects space adversely more than time, whereas, a checkpoint interval that is too large affects time adversely more than space.

Dynamic checkpointing procedure for the design of stabilizing protocols

In this paper, the problem of designing stabilizing computer communication protocols is addressed. A communication protocol is said to be stabilizing, if starting from or being at any illegal global state, the protocol will eventually reach a legal (or consistent) global state, and resume its normal execution. To achieve protocol stabilization, the protocol must be able to detect the error when it occurs, and then it must recover from that error and revert to a legal protocol state. Based on the concepts of event indices and maximally reachable event index tuples, we propose a novel approach for distributed dynamic checkpointing in which the overhead associated with the more traditional periodic checkpointing techniques is avoided. Furthermore, our checkpointing technique can be used as the basis for optimal protocol recovery to achieve stabilization. An example illustrating the new dynamic checkpointing technique is also provided.

An analytical comparison of periodic checkpointing and incremental state saving

The successful application of optimistic synchronization techniques in parallel simulation requires that rollback overheads be contained. The chief contributions to rollback overhead in a Time Warp simulation are the time required to save state information and the time required to restore a previous state. Two competing techniques for reducing rollback overhead are periodic checkpointing (Lin and Lazowska, 1989) and incremental state saving (Bauer et al., 1991). This paper analytically compares the relative performance of periodic checkpointing to incremental state savings. The analytical model derived for periodic checkpointing is based almost entirely on the previous model developed by Lin (Lin and Lazowska, 1989). The analytical model for incremental state saving has been developed for this study. The comparison assumes an optimal checkpoint interval and shows under what simulation parameters each technique performs best.

Performance evaluation of ephemeral logging

Ephemeral logging (EL) is a new technique for managing a log of database activity on disk. It does not require periodic checkpoints and does not abort lengthy transactions as frequently as traditional firewall logging for the same amount of disk space. Therefore, it is well suited for highly concurrent databases and applications which have a wide distribution of transaction lifetimes. This paper briefly explains EL and then analyzes its performance. Simulation studies indicate that it can offer significant savings in disk space, at the expense of slightly higher bandwidth for logging and more main memory. The reduced size of the log implies much faster recovery after a crash as well as cost savings. EL is the method of choice in some but not all situations. We assess the limitations of our current knowledge about EL and suggest promising directions for further research.

Availability of a distributed computer system with failures

A model for distributed systems with failing components is presented. Each node may fail and during its recovery the load is distributed to other nodes that are up. The model assumes periodic checkpointing for error recovery and testing of the status of other nodes for the distribution of load. We consider the availability of a node, which is the proportion of time a node is available for processing, as the performance measure. A methodology for optimizing the availability of a node with respect to the checkpointing and testing intervals is given. A decomposition approach that uses the steady-state flow balance condition to estimate the load at a node is proposed. Numerical examples are presented to demonstrate the usefulness of the technique. For the case in which all nodes are identical, closed form solutions are obtained.

On the availability of a distributed computer system with failing components

We present a model for distributed systems with failing components. Each node may fail and during its recovery the load is distributed to other nodes that are operational. The model assumes periodic checkpointing for error recovery and testing of the status of other nodes for the distribution of load. We consider the availability of a node, which is the proportion of time a node is available for processing, as the performance measure. A methodology for optimizing the availability of a node with respect to the checkpointing and testing intervals is given. A decomposition approach that uses the steady-state flow balance condition to estimate the load at a node is proposed. Numerical examples are presented to demonstrate the usefulness of the technique. For the case in which all nodes are identical, closed form solutions are obtained.

Analysis of a service facility with periodic checkpointing

This paper is an application of the theory of Markov renewal and semi regenerative processes into checkpointing problems. Its main practical contribution consists in the analytic expression of mean response time of systems under checkpointing and in the presence of intermittent failures (data bases, file systems ...).