Real-time transaction processing for autonomic Grid applications

https://doi.org/10.1016/j.engappai.2004.09.002Get rights and content

Abstract

The future Grid will be an autonomic environment that can not only assist users to share large-scale resources and accomplish collaborative tasks but also self-manage to reduce the users’ interventions as much as possible. In such an autonomic Grid environment, the real-time transaction processing is a key and challenging technology to protect systems from various failures. This paper presents an autonomic real-time transaction service (ARTTS) that can (1) dynamically discover Grid services as participants to execute specified sub-transactions, (2) coordinate these participants to achieve the real-time and transactional requirements, and (3) assign priorities to schedule concurrent transactions. Petri nets are used to model and validate the coordination algorithms of real-time Grid transactions and experiment result demonstrates the feasibility of the ARTTS and the performance of the algorithms in different workloads. By handling the potential failures and exceptions autonomically, the ARTTS can facilitate the implementation of real-time Grid transactions and simplify the system management work, which frees users from the complex interference in the autonomic Grid environment.

Introduction

Autonomic computing is a new computing model to self-manage computing systems with a minimal human interference (Autonomic computing, 2004). It provides an unprecedented level of self-regulation and hides complexity from users. The essence of autonomic computing systems is self-management, which focuses on the following four issues (Kephart and Chess, 2003; Bantz et al., 2003):

  • Self-configuration: Autonomic systems automatically configure themselves according to high-level policies. The rest of the systems adjust automatically and seamlessly.

  • Self-optimization: Components and systems continually seek opportunities to improve their own performance and efficiency.

  • Self-healing: These systems automatically detect, diagnose, and repair local software and hardware problems.

  • Self-protection: These systems automatically resist malicious attacks or cascading failures. They use early warnings to anticipate and prevent system-wide failures.

Autonomic Grid computing combines autonomic computing with Grid technologies to help companies harness the collective power of disparate computing resources and manage them as a single, large computer system. Essentially, autonomic Grid computing is a technology for sharing large-scale resources and accomplishing collaborative tasks in a self-configuring, self-optimizing, self-healing, and self-protecting fashion. One goal of autonomic Grid computing is to reduce the complexity associated with the Grid systems to make those systems easy to use and to shield the Grid users from the complex management work as much as possible. Ganek and Corbi (2003) pointed out that a successful Grid system requires autonomic functionality since Grid deployments can expand the domain of computing across many systems. In the autonomic Grid environment, the ability of handling real-time transactions is a key challenging technology because it is an important means to protect systems from the effect of various failures.

In a real-time transaction, the deadline refers to the time by which the transaction must finish or else undesirable results may occur. Based on the strictness of deadlines, real-time Grid transactions can be classified into three kinds, mimicking traditional distributed systems (Kayan and Ulusoy, 1999).

  • Hard real-time transaction: If these transactions miss their deadlines, catastrophic consequences may result.

  • Firm real-time transaction: It is of no value to complete a firm real-time transaction after its deadline has passed. But catastrophic results do not occur after a firm real-time transaction passes its deadline.

  • Soft real-time transaction: Satisfaction of deadline is also the primary performance goal. Unlike a firm real-time transaction, however, there may still be some benefit for completing a soft real-time transaction after its deadline.

This paper proposes an autonomic real-time transaction service (ARTTS), which incorporates fault tolerance into autonomic Grid technologies by automatically recovering systems from various failures. The ARTTS is able to detect and prevent system-wide failures for autonomic Grid systems and handle the entire process of a real-time transaction on behalf of users.

The rest of this paper is organized as follows. The next section reviews some related work. Section 3 introduces the concept of a real-time Grid transaction and modules of the ARTTS. Section 4 presents coordination algorithms of real-time Grid transactions, deadline calculation and priority assignment. Section 5 validates the correctness of coordination algorithms with Petri net models. The experimental result is shown and analyzed in Section 6. Finally, Section 7 summarizes this work.

Section snippets

Service discovery

The first step of handing a Grid transaction is to dynamically discover services to execute sub-transactions. The Universal Description, Discovery, and Integration (UDDI) defines how to publish and discover Web services. Providers of Web services directly publish their services in the UDDI server. Unlike Web services that are persistent, Grid services include both transient and persistent services (Foster et al., 2002). The former refers to the services whose instances are created and/or

Real-time Grid transaction

Tang et al. (2003a) have discussed coordination of different activities in Grid computing and presented corresponding coordination algorithms for two types of transactions, atomic transaction (AT) and cohesion transaction (CT). The AT, served as coordinating the short-lived transaction, consists of a set of atomic sub-transactions that have to commit synchronously. The CT, consisting of atomic sub-transactions or cohesion sub-transactions, allows some sub-transactions to commit while others

Implementation

Transaction processing for the autonomic Grid environment focuses on how to coordinate sub-transactions rather than processing of each individual sub-transaction. The real-time Grid transaction mainly concerns with participant discovery, coordination algorithms, and policies of deadline and priority assignment.

Modeling coordination algorithms with Petri nets

A Petri net is an abstract and formal modeling tool applicable to many systems. It can model systems’ events, conditions and relationships. The occurrence of these events may change the state of the system, causing some of the previous conditions to cease holding and other conditions to begin to hold (Peterson, 1977). Two successful application areas of Petri nets are protocol validation and performance evaluation.

A Petri net is a particular kind of directed graph and consists of two kinds of

Results and discussion

To test the feasibility of the ARTTS and the performance of the algorithms in different workloads, a prototype system has been developed. Each service and its home Local Service Registry reside in the same computer. All services were implemented as transient services. That is, they are instantiated when a consumer discovers their references. All classes were compiled by J2SDK 1.4. In the prototype system, the ARTTS was installed in all computers. The Coordinator and the Participant were Java

Conclusions and future work

This paper has presented an autonomic real-time transaction service ARTTS which can dynamically discover Grid services to execute specified sub-transactions, coordinate these services to achieve the real-time and transactional requirements, and assign priority for scheduling concurrent transactions. As a result, by handling the entire transaction process on behalf of users autonomically, the ARTTS can facilitate the implementation of real-time and transactional Grid applications to provide

Acknowledgments

The authors would like to thank Professor Huaglory Tianfield, Professor Rainer Unland and the reviewers of this paper for their constructive comments and suggestions.

Feilong Tang is a Ph.D. candidate in Department of Computer Science and Engineering, Shanghai Jiao Tong University, China. His research interests focus on Grid computing, Web Services, and computer network.

References (23)

  • R. Abbott et al.

    Scheduling Real-Time Transaction

    ACM SIGMOD Record

    (1988)
  • Autonomic computing, 2004....
  • D.F. Bantz et al.

    Autonomic personal computing

    IBM Systems Journal

    (2003)
  • Cabrera, F., Copel, G., Coxetal, B., 2002a. Web Services Transaction (WS-Transaction)....
  • Cabrera, F., Copel, G., Freundetal, T., 2002b. Web Services Coordination (WS-Coordination)....
  • Ceponkus, A., Cox, W., Brown, G., et al., 2002. Business Transaction Protocol V1.0....
  • Chen, H.R., Chin, Y.H., 2002. An efficient real-time scheduler for nested transaction models. Proceedings of the Ninth...
  • Chen, H.R., Chin, Y.H., Tseng, S. M., 2001. Scheduling value-based transactions in distributed real-time database...
  • Chen, Y.W., Gruenwald, L., 1994. Research issues for a real-time nested transaction model. Proceedings of the IEEE...
  • Foster, I., Kesselman, C., Nick, J.M., et al., 2002. The Physiology of the Grid-An Open Grid Services Architecture for...
  • A.G. Ganek et al.

    The dawning of the autonomic computing era

    IBM Systems Journal

    (2003)
  • Cited by (34)

    • Load balanced scheduling and reliability modeling of grid transaction processing system using colored Petri nets

      2019, ISA Transactions
      Citation Excerpt :

      while scheduling the tasks of processes/jobs with load balancing and without load balancing in different distributed computing environments including the grid system. In the GTP system, a transaction arrives at any node and may execute at several nodes before it commits [18–20]. Due to dynamic behavior of grid computing system, some nodes are added while some nodes are forced to leave the system.

    • On maximizing reliability of grid transaction processing system considering balanced task allocation using social spider optimization

      2018, Swarm and Evolutionary Computation
      Citation Excerpt :

      In such a platform, transaction processing plays a vital role. In grid transaction processing system, the transactions arrive at any node and may execute at several nodes before committing [4–6]. Due to dynamic nature of grid computing system, some nodes add while some nodes leave the system.

    • On scheduling transactions in a grid processing system considering load through Ant Colony Optimization

      2017, Applied Soft Computing
      Citation Excerpt :

      It refers to a reliable and coherent process unit which interacts with one or more systems, independently of the other. It is one of such applications which is widely needed as an effective means by sharing a large number of resources among different organizations [4–6]. A transaction can be categorized as a short-lived transaction which consists of operations representing short-running computations on the data volumes, and a long-lived transaction which consists of operations representing long-running computations on the data volumes [3].

    • Artificial neural networks for short-term load forecasting in microgrids environment

      2014, Energy
      Citation Excerpt :

      SGs are autonomic environments which not only can assist in the coordination of resources to accomplish collaborative tasks but are also capable of self-management to reduce operators' interventions as much as possible. Tang et al. [2] present for instance an autonomic real-time transaction service that can dynamically discover grid services as participants to execute specified sub-transactions, coordinate these participants to achieve the real-time and transactional requirements, and assign priorities to schedule concurrent transactions. According to the Consortium for Electric Reliability Technology Solutions (CERTS), a microgrid is: “… an aggregation of loads and microsources operating as a single system providing both power and heat (… )” [3].

    • An efficient deadlock prevention approach for service oriented transaction processing

      2012, Computers and Mathematics with Applications
      Citation Excerpt :

      Service oriented architecture (SOA) presents new requirements and challenges to the transaction management. With the success of SOA, many large-scale information systems have been set up to provide business services simultaneously [2,3]. In SOA environments, the deadlock problem due to the resource competition among multiple (sub)transactions gets worse because of the randomicity of transaction requests and the uncontrollability of transaction execution order [4,5].

    View all citing articles on Scopus

    Feilong Tang is a Ph.D. candidate in Department of Computer Science and Engineering, Shanghai Jiao Tong University, China. His research interests focus on Grid computing, Web Services, and computer network.

    Minglu Li is a full professor and associate director in Department of Computer Science and Engineering, Shanghai Jiao Tong University, China. Now he also is a director of Grid computing center and Web Services research center of Shanghai Jiao Tong University, Grid expert of Ministry of Education, P.R.China, and expert-in-chief of ShanghaiGrid project. His research interests mainly include Grid computing, Web Services, and multimedia computing.

    Joshua Zhexue Huang is the Assistant Director of the E-Business Technology Institute (ETI) and Honorary Professor of the Department of Mathematics, The University of Hong Kong. He received his Ph.D. degree from The Royal Institute of Technology in Sweden in 1993. From 1994 to 1998, he was a research scientist at CSIRO, Australia, working in the areas of database systems, spatial information systems and data mining. Before joining ETI in 2000, he worked 3 years at MIP Australia as a senior consultant to help Australia companies to implement business intelligence solutions. His current research interests include data mining algorithms, text mining, knowledge-based business intelligence systems and business intelligence service Grid.

    This work is supported by the 973 program of China (No.2002CB312002), National Natural Science Foundation (No. 60473092) of China, ChinaGrid program of MOE of China and the ShanghaiGrid project from the Science and Technology Commission of Shanghai Municipality (No.03dz15027).

    View full text