Charles Zhang

Resolving feature convolution in middleware systems

Middleware provides simplicity and uniformity for the development of distributed applications. However, the modularity of the architecture of middleware is starting to disintegrate and to become complicated due to the interaction of too many orthogonal concerns imposed from a wide range of application requirements. This is not due to bad design but rather due to the limitations of the conventional architectural decomposition methodologies. We introduce the principles of horizontal decomposition (HD) which addresses this problem with a mixed-paradigm middleware architecture. HD provides guidance for the use of conventional decomposition methods to implement the core functionalities of middleware and the use of aspect orientation to address its orthogonal properties. Our evaluation of the horizontal decomposition principles focuses on refactoring major middleware functionalities into aspects in order to modularize and isolate them from the core architecture. New versions of the middleware platform can be created through combining the core and the flexible selection of middleware aspects such as IDL data types, the oneway invocation style, the dynamic messaging style, and additional character encoding schemes. As a result, the primary functionality of the middleware is supported with a much simpler architecture and enhanced performance. Moreover, customization and configuration of the middleware for a wide-range of requirements becomes possible.

LEAP: lightweight deterministic multi-processor replay of concurrent java programs

The technique of deterministic record and replay aims at faithfully reenacting an earlier program execution. For concurrent programs, it is one of the most important techniques for program understanding and debugging. The state of the art deterministic replay techniques face challenging efficiency problems in supporting multi-processor executions due to the unoptimized treatment of shared memory accesses. We propose LEAP: a deterministic record and replay technique that uses a new type of local order w.r.t. the shared memory locations and concurrent threads. Compared to the related work, our technique records much less information without losing the replay determinism. The correctness of our technique is underpinned by formal models and a replay theorem that we have developed in this work. Through our evaluation using both benchmarks and real world applications, we show that LEAP is more than 10x faster than conventional global-order based approaches and, in most cases, 2x to 10x faster than other local-order based approaches. Our recording overhead on the two large open source multi-threaded applications Tomcat and Derby is less than 10%. Moreover, as the evidence of the deterministic replay, LEAP is able to deterministically reproduce 7 out of 8 real bugs in Tomcat and Derby, 13 out of 16 benchmark bugs in IBM ConTest benchmark suite, and 100% of the randomly injected concurrency bugs.

CLAP: recording local executions to reproduce concurrency failures

We present CLAP, a new technique to reproduce concurrency bugs. CLAP has two key steps. First, it logs thread local execution paths at runtime. Second, offline, it computes memory dependencies that accord with the logged execution and are able to reproduce the observed bug. The second step works by combining constraints from the thread paths and constraints based on a memory model, and computing an execution with a constraint solver. CLAP has four major advantages. First, logging purely local execution of each thread is substantially cheaper than logging memory interactions, which enables CLAP to be efficient compared to previous approaches. Second, our logging does not require any synchronization and hence with no added memory barriers or fences; this minimizes perturbation and missed bugs due to extra synchronizations foreclosing certain racy behaviors. Third, since it uses no synchronization, we extend CLAP to work on a range of relaxed memory models, such as TSO and PSO, in addition to sequential consistency. Fourth, CLAP can compute a much simpler execution than the original one, that reveals the bug with minimal thread context switches. To mitigate the scalability issues, we also present an approach to parallelize constraint solving, which theoretically scales our technique to programs with arbitrary execution length. Experimental results on a variety of multithreaded benchmarks and real world concurrent applications validate these advantages by showing that our technique is effective in reproducing concurrency bugs even under relaxed memory models; furthermore, it is significantly more efficient than a state-of-the-art technique that records shared memory dependencies, reducing execution time overhead by 45% and log size by 88% on average.

QSynth: A Tool for QoS-aware Automatic Service Composition

With the proliferation of Web services, service engineers demand good automatic service composition algorithms that not only synthesize the correct work plans from thousands of services but also satisfy the quality requirements of the users. Our observation is that conventional approaches suffer from serious limitations in scalability and accuracy when addressing both requirements simultaneously. We have designed and implemented a tool QSynth to use QoS objectives of service requests as the search directives. This approach effectively prunes the search space and significantly improves the accuracy of the search results. Evaluations show that, compared to the state of the art, QSynth achieves superior scalability and accuracy with respect to a large variety of composition scenarios. Our design of QSynth won the performance championship of Web Services Challenge 2009.

Axis: automatically fixing atomicity violations through solving control constraints

Atomicity, a general correctness criterion in concurrency programs, is often violated in real-world applications. The violations are difficult for developers to fix, making automatic bug fixing techniques attractive. The state of the art approach aims at automating the manual fixing process but cannot provide any theoretical reasoning and guarantees. We provide an automatic approach that applies well-studied discrete control theory to guarantee deadlocks are not introduced and maximal preservation of the concurrency of the original code. Under the hood, we reduce the problem of violation fixing to a constraint solving problem using the Petri net model. Our evaluation on 13 subjects shows that the slowdown incurred by our patches is only 40% of that of the state of the art. With the deadlock-free guarantee, our patches incur moderate overhead (around 10%), which is a worthwhile cost for safety.

Towards just-in-time middleware architectures

Middleware becomes increasingly important in building distributed applications. Today, conventional middleware systems are designed, implemented, and packaged prior to their applications. We argue that with this middleware construction paradigm it is often difficult to meet the challenges imposed by application specific customization requirements. We propose to reverse this paradigm by automatically synthesizing middleware structures as the result of reasoning about the distribution needs of the user application of middleware. We term this type of post-postulated middleware Just-in-time middleware (JiM). In this paper, we present our initial design and present an evaluation of the JiM paradigm through Abacus, a CORBA middleware implementation based on the aspect oriented refactoring of an industrial strength object request broker. In addition, we present Arachne, the Abacus synthesizer, which integrates source analysis, feature inference, and implementation synthesis. Our evaluations show that, through automatic synthesis alone, Abacus is able to support diversified application domains with very flexible architectural compositions and versatile resource requirements as compared to conventional pre-postulated approaches.

PRISM is research in aSpect mining

Aspect oriented programming (AOP) provides programmers with a new level of modularization capability. Although improving the aspect oriented languages seems to be the most eminent task, we recognize that it is equally important for (professional) software developers to properly leverage the power of AOP and to apply AOP at large scales. As many researchers have realized, one of the best ways to assess the practicality of AOP is to retrofit AOP onto legacy software systems [6, 7, 1]. One of the important activities in doing so is aspect mining. Aspect mining aims at finding non-localized programming concerns of very large software systems, which in turn provide opportunities for aspect oriented modularization and aspect oriented refactoring. Existing work of aspect mining focuses on aspect visualization and aspect browsing [3, 2]. These tools often involve tedious human interactions for complex pattern definition over large code base. Inspired by AMT [3], AMTEX [5] is a step towards building an aspect mining environment which facilitates either autonomous or assisted aspect discovery. Prism is a major extension of AMTEX, which is designed for IDE integration, cross-language mining support, and multimodal analysis. Prism is also an extensible mining platform, which is open for supporting new mining techniques. The current implementation of Prism is built as an Eclipse plugin. It supports Java systems and partially supports C# programs. Through Prism, users can define complex lexical, semantic, and syntactical patterns. The Prism engine

Generic middleware substrate through modelware

Conventional middleware architectures suffer from insufficient module-level reusability and the ability to adapt in face of functionality evolution and diversification. To overcome these deficiencies, we propose the Modelware methodology adopting the Model Driven Architecture (MDA) approach and aspect oriented programming (AOP). We advocate the use of models and views to separate intrinsic functionalities of middleware from extrinsic ones. This separation effectively lowers the concern density per component and fosters the coherence and the reuse of the components of middleware architectures. Comparing to the conventionally designed version, Modelware improves the standard benchmark performance by as much as 40% through architectural optimizations. Our evaluation also shows that Modelware considerably reduces coding efforts in supporting the funcitonal evolution of middleware and dramatically different application domains.

Grail: context-aware fixing of concurrency bugs

Writing efficient synchronization for multithreaded programs is notoriously hard. The resulting code often contains subtle concurrency bugs. Even worse, many bug fixes introduce new bugs. A classic example, seen widely in practice, is deadlocks resulting from fixing of an atomicity violation. These complexities have motivated the development of automated fixing techniques. Current techniques generate fixes that are typically conservative, giving up on available parallelism. Moreover, some of the techniques cannot guarantee the correctness of a fix, and may introduce deadlocks similarly to manual fix, whereas techniques that ensure correctness do so at the expense of even greater performance loss. We present Grail, a novel fixing algorithm that departs from previous techniques by simultaneously providing both correctness and optimality guarantees. Grail synthesizes bug-free yet optimal lock-based synchronization. To achieve this, Grail builds an analysis model of the buggy code that is both contextual, distinguishing different aliasing contexts to ensure efficiency, and global, accounting for the entire synchronization behavior of the involved threads to ensure correctness. Evaluation of Grail on 12 bugs from popular codebases confirms its practical advantages, especially compared with existing techniques: Grail patches are, in general, >=40% more efficient than the patches produced by other techniques, and incur only 2% overhead.

Stride: search-based deterministic replay in polynomial time via bounded linkage

Deterministic replay remains as one of the most effective ways to comprehend concurrent bugs. Existing approaches either maintain the exact shared read-write linkages with a large runtime overhead or use exponential off-line algorithms to search for a feasible interleaved execution. In this paper, we propose Stride, a hybrid solution that records the bounded shared memory access linkages at runtime and infers an equivalent interleaving in polynomial time, under the sequential consistency assumption. The recording scheme eliminates the need for synchronizing the shared read operations, which results in a significant overhead reduction. Comparing to the previous state-of-the-art approach of deterministic replay, Stride reduces, on average, 2.5 times of runtime overhead and produces, on average, 3.88 times smaller logs.