Computation Slicing: Techniques and Theory

Abstract

Abstract. We generalize the notion of slice introduced in our earlier paper [6]. A slice of a distributed computation with respect to a global predicate is the smallest computation that contains all consistent cuts of the original computation that satisfy the predicate. We prove that slice exists for all global predicates. We also establish that it is, in general, NP-complete to compute the slice. An optimal algorithm to compute slices for special cases of predicates is provided. Further, we present an efficient algorithm to graft two slices, that is, given two slices, either compute the smallest slice that contains all consistent cuts that are common to both slices or compute the smallest slice that contains all consistent cuts that belong to at least one of the slices. We give application of slicing in general and grafting in particular to global property evaluation of distributed programs. Finally, we show that the results pertaining to consistent global checkpoints [14,18] can be derived as special cases of computation slicing.

Modeling and Analyzing Periodic Distributed Computations

“…In this paper, we instead use directed graphs to model distributed computations as done in [9]. When the graph is acyclic, it represents a distributed computation.…”

“…A lot of work has been done in identifying the classes of predicates which can be efficiently detected [7,9]. However, most of the previous work in this area is mainly restricted to finite traces.…”

“…Some examples of the predicates for which the predicate detection can be solved efficiently are: conjunctive [7,20], disjunctive [7], observer-independent [21,7], linear [7], non-temporal regular [22,9] predicates and temporal [8,23,24].…”

“…We show in this paper that it is sufficient to unroll the d-diagram N times where N is the number of processes in the system. We note here that there has been earlier work in detection of temporal logic formulas on distributed computation, such as [7][8][9][10]. However, the earlier work was restricted to verifying the temporal logic formula on the finite computation with the interpretation of liveness predicates modified to work for finite posets.…”

Monitoring Distributed Controllers: When an Efficient LTL Algorithm on Sequences Is Needed to Model-Check Traces

“…They present an efficient procedure for computing such slices. Computation slicing on regular predicates is examined in details in [MG01]. In [SG03], A. Sen and Garg present the temporal logic RCTL (for regular-CTL), which is a subset of the temporal logic CTL (and an extension, RCTL+).…”