Change-Enabled Software Systems

Few real software systems are built completely from scratch nowadays. Instead, systems are built iteratively and incrementally, while integrating and interacting with components from many other systems. Adaptation, reconfiguration and evolution are normal, ongoing processes throughout the lifecycle of a software system. Nevertheless the platforms, tools and environments we use to develop software are still largely based on an outmoded model that presupposes that software systems are closed and will not significantly evolve after deployment. We claim that in order to enable effective and graceful evolution of modern software systems, we must make these systems more amenable to change by (i) providing explicit, first-class models of software artifacts, change, and history at the level of the platform, (ii) continuously analysing static and dynamic evolution to track emergent properties, and (iii) closing the gap between the domain model and the developers' view of the evolving system. We outline our vision of dynamic, evolving software systems and identify the research challenges to realizing this vision.

Model-Centric, Context-Aware Software Adaptation

Software must be constantly adapted to changing requirements. The time scale, abstraction level and granularity of adaptations may vary from short-term, fine-grained adaptation to long-term, coarse-grained evolution. Fine-grained, dynamic and context-dependent adaptations can be particularly difficult to realize in long-lived, large-scale software systems. We argue that, in order to effectively and efficiently deploy such changes, adaptive applications must be built on an infrastructure that is not just model-driven, but is both model-centric and context-aware. Specifically, this means that high-level, causally-connected models of the application and the software infrastructure itself should be available at run-time, and that changes may need to be scoped to the run-time execution context. We first review the dimensions of software adaptation and evolution, and then we show how model-centric design can address the adaptation needs of a variety of applications that span these dimensions. We demonstrate through concrete examples how model-centric and context-aware designs work at the level of application interface, programming language and runtime. We then propose a research agenda for a model-centric development environment that supports dynamic software adaptation and evolution.

Language Boxes: Bending the Host Language with Modular Language Changes

As domain-specific modeling begins to attract widespread acceptance, pressure is increasing for the development of new domain-specific languages. Unfortunately these DSLs typically conflict with the grammar of the host language, making it difficult to compose hybrid code except at the level of strings; few mechanisms (if any) exist to control the scope of usage of multiple DSLs; and, most seriously, existing host language tools are typically unaware of the DSL extensions, thus hampering the development process. Language boxes address these issues by offering a simple, modular mechanism to encapsulate (i) compositional changes to the host language, (ii) transformations to address various concerns such as compilation and highlighting, and (iii) scoping rules to control visibility of language extensions. We describe the design and implementation of language boxes, and show with the help of several examples how modular extensions can be introduced to a host language and environment.

Enriching Reverse Engineering with Annotations

Much of the knowledge about software systems is implicit, and therefore difficult to recover by purely automated techniques. Architectural layers and the externally visible features of software systems are two examples of information that can be difficult to detect from source code alone, and that would benefit from additional human knowledge. Typical approaches to reasoning about data involve encoding an explicit meta-model and expressing analyses at that level. Due to its informal nature, however, human knowledge can be difficult to characterize up-front and integrate into such a meta-model. We propose a generic, annotation-based approach to capture such knowledge during the reverse engineering process. Annotation types can be iteratively defined, refined and transformed, without requiring a fixed meta-model to be defined in advance. We show how our approach supports reverse engineering by implementing it in a tool called Metanool and by applying it to (i) analyzing architectural layering, (ii) tracking reengineering tasks, (iii) detecting design flaws, and (iv) analyzing features.

Practical Object-Oriented Back-in-Time Debugging

Back-in-time debuggers are extremely useful tools for identifying the causes of bugs, as they allow us to inspect the past states of objects no longer present in the current execution stack. Unfortunately the "omniscient" approaches that try to remember all previous states are impractical because they either consume too much space or they are far too slow. Several approaches rely on heuristics to limit these penalties, but they ultimately end up throwing out too much relevant information. In this paper we propose a practical approach to back-in-time debugging that attempts to keep track of only the relevant past data. In contrast to other approaches, we keep object history information together with the regular objects in the application memory. Although seemingly counter-intuitive, this approach has the effect that past data that is not reachable from current application objects (and hence, no longer relevant) is automatically garbage collected. In this paper we describe the technical details of our approach, and we present benchmarks that demonstrate that memory consumption stays within practical bounds. Furthermore since our approach works at the virtual machine level, the performance penalty is significantly better than with other approaches.

Back to the future in one week - implementing a Smalltalk VM in PyPy

We report on our experiences with the Spy project, including implementation details and benchmark results. Spy is a re-implementation of the Squeak (i.e., Smalltalk-80) VM using the PyPy toolchain. The PyPy project allows code written in RPython, a subset of Python, to be translated to a multitude of different backends and architectures. During the translation, many aspects of the implementation can be independently tuned, such as the garbage collection algorithm or threading implementation. In this way, a whole host of interpreters can be derived from one abstract interpreter definition. Spy aims to bring these benefits to Squeak, allowing for greater portability and, eventually, improved performance. The current Spy codebase is able to run a small set of benchmarks that demonstrate performance superior to many similar Smalltalk VMs, but which still run slower than in Squeak itself. Spy was built from scratch over the course of a week during a joint Squeak-PyPy Sprint in Bern last autumn.

A Calculus of Evolving Objects

The demands of developing modern, highly dynamic applications have led to an increasing interest in dynamic programming languages and mechanisms. Not only must applications evolve over time, but the object models themselves may need to be adapted to the requirements of different run-time contexts. Class-based models and prototype-based models, for example, may need to co-exist to meet the demands of dynamically evolving applications. Multi-dimensional dispatch, fine-grained and dynamic software composition, and run-time evolution of behaviour are further examples of diverse mechanisms which may need to co-exist in a dynamically evolving run-time environment. How can we model the semantics of these highly dynamic features, yet still offer some reasonable safety guarantees? To this end we present an original calculus in which objects can adapt their behaviour at run-time. Both objects and environments are represented by first-class mappings between variables and values. Message sends are dynamically resolved to method calls. Variables may be dynamically bound, making it possible to model a variety of dynamic mechanisms within the same calculus. Despite the highly dynamic nature of the calculus, safety properties are assured by a type assignment system.

Quality of Service for Multicasting in Content Addressable Networks

Multicasting is an efficient mechanism for one to many data dissemination. Unfortunately, IP Multicasting is not widely available to end-users today, but Application Layer Multicast (ALM), such as Content Addressable Network, helps to overcome this limitation. Our OM-QoS framework offers Quality of Service support for ALMs. We evaluated OM-QoS applied to CAN and show that we can guarantee that all multicast paths support certain QoS requirements.

Hop-to-Hop Reliability in IP-Based Wireless Sensor Networks - A Cross-Layer Approach

To interconnect a wireless sensor network (WSN) to the Internet, we propose to use TCP/IP as the standard protocol for all network entities. We present a cross layer designed communication architecture, which contains a MAC protocol, IP, a new protocol called Hop-to-Hop Reliability (H2HR) protocol, and the TCP Support for Sensor Nodes (TSS) protocol. The MAC protocol implements the MAC layer of beacon-less personal area networks (PANs) as defined in IEEE 802.15.4. H2HR implements hop-to-hop reliability mechanisms. Two acknowledgment mechanisms, explicit and implicit ACK are supported. TSS optimizes using TCP in WSNs by implementing local retransmission of TCP data packets, local TCP ACK regeneration, aggressive TCP ACK recovery, congestion and flow control algorithms. We show that H2HR increases the performance of UDP, TCP, and RMST in WSNs significantly. The throughput is increased and the packet loss ratio is decreased. As a result, WSNs can be operated and managed using TCP/IP.

Gravity-based Local Clock Synchronization in Wireless Sensor Networks

Contention-based MAC protocols follow periodic listen/sleep cycles. These protocols face the problem of virtual clustering if different unsynchronized listen/sleep schedules occur in the network, which has been shown to happen in wireless sensor networks. To interconnect these virtual clusters, border nodes maintaining all respective listen/sleep schedules are required. However, this is a waste of energy, if locally a common schedule can be determined. We propose to achieve local synchronization with a mechanism that is similar to gravitation. Clusters represent the mass, whereas synchronization messages sent by each cluster represent the gravitation force of the according cluster. Due to the mutual attraction caused by the clusters, all clusters merge finally. The exchange of synchronization messages itself is not altered by LACAS. Accordingly, LACAS introduces no overhead. Only a not yet used property of synchronization mechanisms is exploited.