Model-Based Adaptation of Software Communicating via FIFO Buffers

Abstract

Abstract. Software Adaptation is a non-intrusive solution for composing black-box components or services (peers) whose individual functionality is as required for the new system, but that present interface mismatch, which leads to deadlock or other undesirable behaviour when combined. Adaptation techniques aim at automatically generating new components called adapters. All the interactions among peers pass through the adapter, which acts as an orchestrator and makes the involved peers work correctly together by compensating for mismatch. Most of the existing solutions in this field assume that peers interact synchronously using rendezvous communication. However, many application areas rely on asynchronous communication models where peers interact exchanging messages via buffers. Generating adapters in this context becomes a difficult problem because peers may exhibit cyclic behaviour, and their composition often results in infinite systems. In this paper, we present a method for automatically generating adapters in asynchronous environments where peers interact using FIFO buffers.

Stability-Based Adaptation of Asynchronously Communicating Software

“…message to the server. Hence, the results in [9] do not apply to our example, whereas the approach presented in this paper works as we will show in the rest of this section.…”

“…In some complex adaptation scenarios, adaptation rules must be taken contextually (i.e., vectors cannot be applied at any time, but only in certain situations). For this purpose, we may use regular expressions (regex) on vectors [9], indicating a pattern for applying them that will constrain the adaptation process, enforcing additional properties on the adapter. This endows adaptation contracts with extended expressivity, though in this work we do not show their use, in order to avoid additional complexity in the presentation of our proposal.…”

“…Then, we can check whether the system is synchronizable. If that were the case, we would be under the conditions defined in [9], which require the asynchronous system to be behaviourally equivalent to its synchronous version. This is not the case of our adapted Client/Server system, in which the client is able to issue several output messages (user!, pwd!, request!)…”

“…In [9], we presented a solution to the software adaptation problem by using the synchronizability property and adapter generation techniques for synchronous communication. The adapter synthesis in this approach relies on an iterative process, which works properly in asynchronous environments.…”

“…We assume that peers are modelled using behavioural descriptions and interact asynchronously via (possibly unbounded) FIFO buffers. In a previous work [9], we presented a preliminary proposal for asynchronous adaptation in which a sufficient condition, called synchronizability, was required. However, many asynchronous systems are not synchronizable.…”