Memory exclusion: optimizing the performance of checkpointing systems

Abstract

Checkpointing systems are a convenient way for users to make their programs fault-tolerant by intermittently saving program state to disk and restoring that state following a failure. The main concern with checkpointing is the overhead that it adds to running time of the program. This paper describes memory exclusion, an important class of optimizations that reduce the overhead of checkpointing. Some forms of memory exclusion are well-known in the checkpointing community. Others are relatively new. In this paper, we describe all of them within the same framework. We have implemented these optimization techniques in two checkpointers: libckpt, which works on Unix-based workstations, and CLIP, which works on the Intel Paragon. Both checkpointers are publicly available at no cost. We have checkpointed various long-running applications with both checkpointers and have explored the performance improvements that may be gained through memory exclusion. Results from these experiments are presented and show the improvements in time and space overhead. Excluding the code segment and using the stack pointerCheckpoints store the address space of a processor. Typically these address spaces have four segments: executable code, global data, heap and stack. In most computer systems, the

On the Viability of Compression for Reducing the Overheads of Checkpoint/Restart-Based Fault Tolerance

“…The second set of strategies reduce commit latencies by reducing checkpoint sizes. These strategies include memory exclusion [19] and incremental checkpointing [20]- [22]. In Section V, we discuss the potential interplay between these optimizations and checkpoint compression.…”

Static analysis and compiler design for idempotent processing

“…Hardware checkpointing resources are also rarely exposed to software, and are even less often configurable in terms of their checkpointing granularity, limiting their wider applicability. Finally, checkpoints have limited application visibility and are often overly aggressive in saving more state than is required by the application [9,27].…”

A fast restart mechanism for checkpoint/recovery protocols in networked environments

“…The other major direction is to reduce checkpoint overhead, especially the disk I/O time. Latency hiding and memory exclusion are two key techniques [16]. The studies in this category include copy-on-write [9], diskless checkpointing [17], and incremental checkpointing [5,20] There also exist several optimization techniques that utilize memory paging mechanisms to achieve fast process execution.…”

“…The effectiveness of FREM requires that the process only access a relatively small portion of its address space within a given time window after a checkpoint. This assumption is justified by two facts in practice: (1) many applications demonstrate good temporal locality in data accesses, and (2) applications using dynamic memory allocation may have a large amount of unused or dead data in their checkpoint image files [16].…”