Weigang Wu

Model Based Verification of Dynamically Evolvable Service Oriented Systems

Dynamic evolution is highly desirable for service oriented systems in open environments. For the evolution to be trusted, it is crucial to keep the process consistent with the specification. In this paper, we study two kinds of evolution scenarios and propose a novel verification approach based on hierarchical timed automata to model check the underlying consistency with the specification. It examines the procedures before, during and after the evolution process, respectively and can support the direct modeling of temporal aspects, as well as the hierarchical decomposition of software structures. Probabilities are introduced to model the uncertainty characterized in open environments and thus can support the verification of parameter-level evolution. We present a flattening algorithm to facilitate automated verification using the mainstream timed automata based model checker–UPPAAL (integrated with UPPAAL-SMC). We also provide a motivating example with performance evaluation that complements the discussion and demonstrates the feasibility of our approach.摘要创新点针对开放环境下面向服务架构的动态演化问题, 文章提出了一种基于扩展的层次式时间自动机的验证方法, 该方法可以对时间相关属性, 层次结构特征, 状态及演化行为进行直接的建模。通过引入概率转移可对开放环境下的不确定性进行建模, 结合应用统计模型检验技术, 从而支持对概率相关规约进行分析验证; 针对结构型的演化类别, 通过传统的时序逻辑进行规约描述, 且可以从演化前、演化中及演化后分别进行验证; 针对参数型的演化类别, 可以采用概率时序逻辑进行规约描述并进行验证, 为说明方法的有效性, 通过一个经典的电子商务系统范例展示了所提方法针对这两种演化类型的验证。

Analyzing Eventual Leader Election Protocols for Dynamic Systems by Probabilistic Model Checking

Leader election protocols have been intensively studied in distributed computing, mostly in the static setting. However, it remains a challenge to design and analyze these protocols in the dynamic setting, due to its high uncertainty, where typical properties include the average steps of electing a leader eventually, the scalability etc. In this paper, we propose a novel model-based approach for analyzing leader election protocols of dynamic systems based on probabilistic model checking. In particular, we employ a leading probabilistic model checker, PRISM, to simulate representative protocol executions. We also relax the assumptions of the original model to cover unreliable channels which requires the introduction of probability to our model. The experiments confirm the feasibility of our approach.

Hierarchical Eventual Leader Election for Dynamic Systems

Dynamic system is a recent hot research topic in theoretical distributed computing. The dynamicity caused by process join and leave bring new challenges in fundamental distributed computing problems, such as eventual leader election. In this paper, we consider leader election in dynamic systems with cluster-based hierarchy. Clustering based hierarchy has been used in fundamental distributed algorithms to achieve scalability and low communication cost, but, to the best of our knowledge, it is not considered in eventual leader election, especially in eventual leader for dynamic systems. We firstly define new system models to describe the dynamicity of clusters, and then based on these models, we design an algorithm to elect an eventual leader. With cluster hierarchy, leader election is basically conducted in two layers. In the lower layer, cluster-heads are elected with each cluster. Then, in the upper layer, election is conducted among cluster-heads so as to elect the eventual leader of the whole system. Several key challenging issues caused by cluster dynamicity have been addressed in our design, including blocking in election within a cluster and multiple cluster-heads in election of upper layer. The proposed algorithm is proved to be correct rigorously.

MRFS: A Distributed Files System with Geo-replicated Metadata

Distributed file system is one of the key blocks of data centers. With the advance in geo-replicated storage systems across data centers, both system scale and user scale are becoming larger and larger. Then, a single metadata server in distributed file system may lead to capacity bottleneck and high latency without considering locality. In this paper, we present the design and implementation of MRFS (Metadata Replication File System), a distributed file system with hierarchical and efficient distributed metadata management, which introduces multiple metadata servers (MDS) and an additional namespace server (NS). Metadata is divided into non-overlapping parts and stored on MDS in which the creation operation is raised, while namespace and directory information is maintained in NS. Such a hierarchical design not only achieves high scalability but also provides low-latency because it satisfies a majority of requests in local MDS. To address hotspot issues and flash crowds, the system supports flexible and configurable metadata replication among MDSs. Evaluation results show that our system MRFS is effective and efficient, and the replication mechanism brings substantial local visit at the cost of affordable memory overhead under various scenarios.

PPMS: A Peer to Peer Metadata Management Strategy for Distributed File Systems

Distributed file system is one of the key blocks of cloud computing systems. With the fast increase of user scale and data amount, metadata management has become a crucial point affecting the overall performance of a distributed file system. In this paper, we design and implement PPMS, a novel metadata management strategy in a peer to peer way. Different from existing metadata management methods, we adopt a two layer structure to achieve high scalability and low latency. The upper layer is metadata index server, which is used to store metadata of directories, while the lower layer consists of metadata servers to store the metadata of files. More importantly, the lower layer is organized in a peer to peer way to further improve scalability. We implement a prototype file system based on PPMS and evaluate its performance via experiments. The results show that our design can achieve high performance with in terms of time latency and system throughput.

Probabilistic verification of hierarchical leader election protocol in dynamic systems

Leader election protocols are fundamental for coordination problems—such as consensus—in distributed computing. Recently, hierarchical leader election protocols have been proposed for dynamic systems where processes can dynamically join and leave, and no process has global information. However, quantitative analysis of such protocols is generally lacking. In this paper, we present a probabilistic model checking based approach to verify quantitative properties of these protocols. Particularly, we employ the compositional technique in the style of assume-guarantee reasoning such that the sub-protocols for each of the two layers are verified separately and the correctness of the whole protocol is guaranteed by the assume-guarantee rules. Moreover, within this framework we also augment the proposed model with additional features such as rewards. This allows the analysis of time or energy consumption of the protocol. Experiments have been conducted to demonstrate the effectiveness of our approach.

Consensus in Smart Computing Systems

A smart computing system is usually composed of smart mobile nodes, such as mobile phones and/or sensor nodes. In this paper, we study the consensus problem in a smart computing system, which has many applications in smart computing, such as smart grid. Due to mobility and failures, smart computing nodes may join and leave a system from time to time. Such dynamicity in node set of a smart computing system brings new challenges in consensus protocol design. A process newly joining the system may miss messages sent out before the joining, which will cause existing consensus algorithms blocked forever and cannot terminate. Process leave may cause the value of previous decision lost, which will make existing consensus algorithms decide on multiple values and violate the agreement property. To address such challenges, we firstly define necessary dynamicity model to constrain the change of process set, and then design a dynamic consensus algorithm which can handle process join and leave effectively via two novel mechanisms. The first mechanism lets old processes send redeeming messages to new processes, while the second one lets processes jump to higher rounds under some specific scenarios. The correctness of our proposed algorithm is rigorously proved, in terms of termination, agreement and validity.

DTC: A Dynamic Transaction Chopping Technique for Geo-replicated Storage Systems

Large Web applications usually require replicating data across geo-distributed datacenters to achieve high locality, durability and availability. However, maintaining strong consistency in geo-replicated systems usually suffers from long latency due to costly coordination across datacenters. Among others, transaction chopping is an effective and efficient approach to cope with such a challenge. In this paper, we propose DTC (Dynamic Transaction Chopping), a novel technique that chops transactions and checks their conflicts in a dynamic and automatic way, during application execution. DTC mainly consists of two parts: a dynamic chopper that chops transaction dynamically according to data partition scheme, and a conflict detection algorithm for determining the safety of the dynamic chopping. Compared with existing transaction chopping technique for geo-replicated systems, DTC has several advantages, including transparency to programmers, flexibility in conflict analysis, high degree of piecewise execution, and adaptability to dynamic partition schemes. We implement our DTC technique and conduct experiments to examine the correctness of DTC and evaluate its performance. The experiment results show that our DTC technique can achieve much more piecewise execution than the existing chopping approach does, and reduce execution time obviously.

Road Segment Information Based Named Data Networking for Vehicular Environments

Vehicular Ad Hoc Network (VANET) can provide strong support for intelligent transportation systems and Internet services. How to disseminate data to vehicular nodes has been always at the core of VANET technologies. In this paper, we consider data dissemination via Named Data Networking (NDN), which is a new and promising Internet technology to realize content centric networking. NDN can route and forward data according to data content (or ID) rather than the address/location of the data, so that it can disseminate and share data more efficiently. Our work focuses on how to cope with the mobility of vehicular nodes, which is not considered in original NDN. Different from existing NDN for vehicular environments, we propose to establish data dissemination route using road segment information rather than node information. Such a new approach can avoid the effect of topology changes on data routes so as to keep data routes stable. A road segment route is instantiated as routes of nodes upon data forwarding demands. The mechanism to realize such an instantiation is the major challenge addressed in our design. Simulations via ndnSIM clearly show that our design performs much better than existing ones.

Processing Partially Ordered Requests in Distributed Stream Processing Systems

In many application scenarios of distributed stream processing, there might be partial order relations among the requests. However, existing stream processing systems can not directly handle partially ordered requests, while indirect mechanisms are usually strongly coupled with business logic, which lack flexibility and have limited performance. We propose Pork, a novel distributed stream processing system targeting at partially ordered requests. In the experiments, the new system has achieved a parallelism and request throughput larger than the traditional mechanism in the presented example, and the performance overhead due to parallelism is considerably small. Then the scalability characteristic of the new system is discussed. What’s more, the experiment results also show that the new system has a more flexible load balancing ability.