On Compensation Primitives as Adaptable Processes

Abstract

We compare mechanisms for compensation handling and dynamic update in calculi for concurrency. These mechanisms are increasingly relevant in the specification of reliable communicating systems. Compensations and updates are intuitively similar: both specify how the behavior of a concurrent system changes at runtime in response to an exceptional event. However, calculi with compensations and updates are technically quite different. We investigate the relative expressiveness of these calculi: we develop encodings of core process languages with compensations into a calculus of adaptable processes developed in prior work. Our encodings shed light on the (intricate) semantics of compensation handling and its key constructs. They also enable the transference of existing verification and reasoning techniques for adaptable processes to core languages with compensation handling.

Efficient compensation handling via subjective updates

“…More precisely, our prior work [7] does not address the following question: to what extent the direction of movement in (1) is responsible for the complexity and cost of the encodings? To clarify this issue, in this paper we encode compensation primitives into adaptable processes with subjective update.…”

“…Our prior work [7] encoded the six calculi in [9] into adaptable processes with objective update as in Rule (1). These encodings, however, are not as efficient as one may like: objective updates turn out to be inconvenient when "collecting" protected blocks scattered within nested transactions.…”

“…These encodings, however, are not as efficient as one may like: objective updates turn out to be inconvenient when "collecting" protected blocks scattered within nested transactions. Roughly, the problem is that objective update leaves processes in the "wrong location"; as a result, the encodings in [7] need additional adaptation steps to bring processes into appropriate locations and achieve the intended semantics. This difficulty reflects prominently in the number of reductions required to mimic a compensation step.…”

“…This difficulty reflects prominently in the number of reductions required to mimic a compensation step. That is, the encodings in [7] are inefficient because they are costly in terms of computational steps. Even if the encodings in [7] add only a few reductions per each compensation, the overall effect of such additional steps must be analyzed in the context of systems with many (nested) transactions, for which the cost of extra reduction steps may rapidly escalate.…”

“…To clarify this issue, in this paper we encode compensation primitives into adaptable processes with subjective update. This is a form of dynamic update not studied in [7], which implements the opposite direction of process movement implemented by objective update. Indeed, in subjective update a located process moves to a (remote) context containing an update process:…”