Implementation and Evaluation of a Protocol for Recording Process Documentation in the Presence of Failures

Abstract

Abstract. The provenance of a particular data item is the process that led to that piece of data. Previous work has enabled the creation of detailed representation of past executions for determining provenance, termed process documentation. However, current solutions to recording process documentation assume a failure free environment. Failures result in process documentation not being recorded, thereby causing the loss of evidence that a process occurred. We have designed F-PReP, a protocol to guarantee the recording of process documentation in the presence of failures. This paper discusses its implementation and evaluates its performance. The result reveals that it introduces acceptable overhead.

Automated and non-intrusive provenance capture with UML2PROV

“…Our evaluation has shown that the SetBindingsConfiguration has the lowest performance penalty. This strategy, based on a bulk submission of bindings, is aligned with other proposals [7,16] that lead to savings in the overhead of establishing extra database connections. This configuration also follows the more compact storage approach, although, it is not the most convenient option from the final provenance consumer point of view (the PROV document is not directly available).…”

Recording Process Documentation in the Presence of Failures

“…Its performance has been evaluated and the result reveals that it introduces acceptable overhead [7]. We are currently investigating how to create process documentation when an application has its own fault tolerance schemes to tolerate application level failures.…”

Why workflows break — Understanding and combating decay in Taverna workflows

“…• Repair by provenance support: Sometimes unavailable services can neither be substituted nor repaired. Provenance information can be useful to explain failures and to enable appropriate repairs, by tracing what changes to the components of a workflow led to this incapability of replicating [10], [13], [23]. Additionally, provenance information can also be used to run the workflow using previous states of the underlying data sources, thereby guaranteeing the reproducibility of the workflow results [24] Approaches like Virtual Machines have also been exploited [34] to preserve runtime environment to cope with changes.…”