"... In PAGE 4: ... From the above discussion, there are contradicting de- mands on the checkpoint interval by I/O performance and error coverage requirements. Table1 summarizes the oppos- ing factors when trying to balance I/O performance, error coverage, and comparison bandwidth in choosing a DMR de- tection mechanisms. Only ngerprinting, presented in detail in the next section, can simultaneously satisfy all require- ments of high error coverage, low detection bandwidth, and high I/O throughput.... ..."

"... In PAGE 4: ... From the above discussion, there are contradicting de- mands on the checkpoint interval by I/O performance and error coverage requirements. Table1 summarizes the oppos- ing factors when trying to balance I/O performance, error coverage, and comparison bandwidth in choosing a DMR de- tection mechanisms. Only ngerprinting, presented in detail in the next section, can simultaneously satisfy all require- ments of high error coverage, low detection bandwidth, and high I/O throughput.... ..."

"... In PAGE 28: ...4 Comparison Various rollback-recovery protocols offer different tradeoffs with respect to perfor- mance overhead, latency of output commit, storage overhead, ease of garbage col- lection, simplicity of recovery, freedom from domino effect, freedom from orphan processes, and the extent of rollback. Table1 summarizes the comparison between the different variations of rollback-recovery protocols. Uncoordinated checkpointing generally has the lowest failure-free overhead but suffers from potential domino ef- fect.... ..."

"... In PAGE 28: ...4 Comparison Various rollback-recovery protocols offer different tradeoffs with respect to perfor- mance overhead, latency of output commit, storage overhead, ease of garbage col- lection, simplicity of recovery, freedom from domino effect, freedom from orphan processes, and the extent of rollback. Table1 summarizes the comparison between the different variations of rollback-recovery protocols. Uncoordinated checkpointing generally has the lowest failure-free overhead but suffers from potential domino ef- fect.... ..."

"... In PAGE 28: ...4 Comparison Various rollback-recovery protocols offer different tradeoffs with respect to perfor- mance overhead, latency of output commit, storage overhead, ease of garbage col- lection, simplicity of recovery, freedom from domino effect, freedom from orphan processes, and the extent of rollback. Table1 summarizes the comparison between the different variations of rollback-recovery protocols. Uncoordinated checkpointing generally has the lowest failure-free overhead but suffers from potential domino ef- fect.... ..."

"... In PAGE 23: ...5 Comparison Different rollback-recovery protocols offer different tradeoffs with respect to properties such as performance overhead, latency of output commit, storage overhead, ease of garbage collection, simplicity of recovery, freedom from domino effect, freedom from orphan processes, and the extent of rollback. Table1 provides a brief comparison between the different styles of rollback-recovery protocols. Since garbage collection and recovery both involve calculating a recovery line, they can be performed by simple procedures under coordinated checkpointing and pessimistic logging, both of which have a predetermined recovery line during failure-free execution.... ..."

"... In PAGE 20: ...5 Comparison Different rollback-recovery protocols offer different tradeoffs with respect to performance overhead, latency of output commit, storage overhead, ease of garbage collection, simplicity of recovery, freedom from domino effect, freedom from orphan processes, and the extent of rollback. Table1 summarizes a comparison between the different variations of rollback-recovery protocols. Since garbage collection and recovery both involve calculating a recovery line, they can be performed by simple procedures under coordinated checkpointing and pessimistic logging, both of which have a predetermined recovery line during failure-free execution.... ..."

"... In PAGE 23: ...5 Comparison Different rollback-recovery protocols offer different tradeoffs with respect to properties such as performance overhead, latency of output commit, storage overhead, ease of garbage collection, simplicity of recovery, freedom from domino effect, freedom from orphan processes, and the extent of rollback. Table1 provides a brief comparison between the different styles of rollback-recovery protocols. Since garbage collection and recovery both involve calculating a recovery line, they can be performed by simple procedures under coordinated checkpointing and pessimistic logging, both of which have a predetermined recovery line during failure-free execution.... ..."

"... In PAGE 20: ...5 Comparison Different rollback-recovery protocols offer different tradeoffs with respect to performance overhead, latency of output commit, storage overhead, ease of garbage collection, simplicity of recovery, freedom from domino effect, freedom from orphan processes, and the extent of rollback. Table1 summarizes a comparison between the different variations of rollback-recovery protocols. Since garbage collection and recovery both involve calculating a recovery line, they can be performed by simple procedures under coordinated checkpointing and pessimistic logging, both of which have a predetermined recovery line during failure-free execution.... ..."

"... In PAGE 20: ...5 Comparison Different rollback-recovery protocols offer different tradeoffs with respect to performance overhead, latency of output commit, storage overhead, ease of garbage collection, simplicity of recovery, freedom from domino effect, freedom from orphan processes, and the extent of rollback. Table1 summarizes a comparison between the different variations of rollback-recovery protocols. Since garbage collection and recovery both involve calculating a recovery line, they can be performed by simple procedures under coordinated checkpointing and pessimistic logging, both of which have a predetermined recovery line during failure-free execution.... ..."