Networks for Grid Applications

Frontmatter

A High Performance SOAP Engine for Grid Computing

Abstract

Web Service technology still has many defects that make its usage for Grid computing problematic, most notably the low performance of the SOAP engine. In this paper, we develop a novel SOAP engine called SOAPExpress, which adopts two key techniques for improving processing performance: SCTP data transport and dynamic early binding based data mapping. Experimental results show a significant and consistent performance improvement of SOAPExpress over Apache Axis.

Ning Wang, Michael Welzl, Liang Zhang

UDTv4: Improvements in Performance and Usability

Abstract

This paper presents UDT version 4 (UDTv4), the fourth generation of the UDT high performance data transfer protocol. The focus of the paper is on the new features introduced in version 4 during the past two years to improve the performance and usability of the protocol.

UDTv4 introduces a new three-layer protocol architecture (connection-flow-multiplexer) for enhanced congestion control and resource management. The new design allows protocol parameters to be shared by parallel connections and to be reused by future connections. This improves the congestion control and reduces the connection setup time. Meanwhile, UDTv4 also provide better usability by supporting a broader variety of network environments and use scenarios.

Yunhong Gu, Robert Grossman

The Measurement and Modeling of a P2P Streaming Video Service

Abstract

Most of the work on grid technology in video area has been generally restricted to aspects of resource scheduling and replica management. The traffic of such service has a lot of characteristics in common with that of the traditional video service. However the architecture and user behavior in Grid networks are quite different from those of traditional Internet. Considering the potential of grid networks and video sharing services, measuring and analyzing P2P IPTV traffic are important and fundamental works in the field grid networks.

This paper investigates the features of PPLive, which is the most popular streaming service in China and based on P2P technology. Through monitoring and analyzing PPLive traffic streams, the characteristics of P2P streaming service have been studied. The analyses are carried out in respect of bearing protocols, geographical distribution and the self-similarity properties of the traffic. A streaming service traffic model has been created and verified with the simulation. The simulation results indicate that the proposed streaming service traffic model complies well with the real IPTV streaming service. It can also function as a step towards studying video-sharing services on grids.

Peng Gao, Tao Liu, Yanming Chen, Xingyao Wu, Yehia El-khatib, Christopher Edwards

SCE: Grid Environment for Scientific Computing

Abstract

Over the last few years Grid computing has evolved into an innovating technology and gotten increased commercial adoption. However, existing Grids do not have enough users as for sustainable development in the long term. This paper proposes several suggestions to this problem on the basis of long-term experience and careful analysis. The Scientific Computing Environment (SCE) in the Chinese Academy of Sciences is introduced as a completely new model and a feasible solution to this problem.

Haili Xiao, Hong Wu, Xuebin Chi

QoS Differentiated Adaptive Scheduled Optical Burst Switching for Grid Networks

Abstract

This paper presents the QoS differentiated Adaptive Scheduled Optical Burst Switching (QAS-OBS) paradigm that efficiently supports dynamic grid applications. (QAS-OBS) paradigm enables differentiated-services optical burst switching via two-way reservation and adaptive QoS control in distribute wavelength-routed networks. Based on the network loading status, the edge nodes dynamically optimize per-class lightpath reservation control to provide proportional QoS differentiation, while maximizing the overall network throughput. Unlike existing QoS control schemes that require special core network to support burst switching and QoS control, QAS-OBS pushes all the complexity of burst switching and QoS control to the edge nodes, so that the core network design is maximally simplified. Simulation results evaluate the performance of QAS-OBS and verified the proposed control schemes could provide proportional QoS control while maximizing the overall network throughput.

Oliver Yu, Huan Xu

Principles of Service Oriented Operating Systems

Abstract

Grid middleware support and the Web Services domain have advanced significantly over recent years, reaching a point where resource exposition and usage over the web has not only become feasible, but an actual commercial reality. Nonetheless, commercial uptake is still slow, and certainly not progressing the same way as other web developments have been taking place - this is mostly due to the fact that usage is still complicated and restrictive. This paper will discuss a new approach towards tackling grid-like networking across organisational boundaries and across different types of resources by moving main parts of the infrastructure to a lower (OS) level. This will allow more intuitive use of Grid infrastructures for current types of users.

Lutz Schubert, Alexander Kipp

Preliminary Resource Management for Dynamic Parallel Applications in the Grid

Abstract

Dynamic parallel applications such as CFD-OG impose a new problem for distributed processing because of their dynamic resource requirements at run-time. These applications are difficult to adapt in the current distributed processing model (such as the Grid) due to a lack of interface for them to directly communicate with the runtime system and the delay of resource allocation. In this paper, we propose a novel mechanism, the Application Agent (AA) embedded between an application and the underlying conventional Grid middleware to support the dynamic resource requests on the fly. We introduce AA’s dynamic process management functionality and its resource buffer policies which efficiently store resources in advance to maintain the execution performance of the application. To this end, we introduce the implementation of AA.

Hao Liu, Amril Nazir, Søren-Aksel Sørensen

Performance Evaluation of a SLA Negotiation Control Protocol for Grid Networks

Abstract

A framework for an autonomous negotiation control protocol for service delivery is crucial to enable the support of heterogeneous service level agreements (SLAs) that will exist in distributed environments. We have first given a gist of our augmented service negotiation protocol to support distinct service elements. The augmentations also encompass related composition of the services and negotiation with several service providers simultaneously. All the incorporated augmentations will enable to consolidate the service negotiation operations for telecom networks, which are evolving towards Grid networks. Furthermore, our autonomous negotiation protocol is based on a distributed multi-agent framework to create an open market for Grid services. Second, we have concisely presented key simulation results of our work in progress. The results exhibit the usefulness of our negotiation protocol for realistic scenarios that involves different background traffic loading, message sizes and traffic flow asymmetry between background and negotiation traffics.

Igor Cergol, Vinod Mirchandani, Dominique Verchere

Performance Assessment Architecture for Grid

Abstract

For the sake of simplicity in resource sharing, Grid services only expose a function access interface to users. However, some Grid users want to know more about the performance of services in planning phase. Problem emerges since the Grid simply has no means in obtaining how well the system will cope with user demands. Current Grid infrastructures do not integrate adequate performance assessment measures to meet the user’s requirement. In this paper, the architecture of Guided Subsystem Approach for Grid performance assessments is presented. We proposed an assessment infrastructure that allows the user to collect information to evaluate the performance of Grid applications. Based on this infrastructure, a user-centric performance assessment method is given. It is expected that this research will lead to some sort of extension in Grid middleware to facilitate the Grid platform the ability to handle applications with higher reliability requirements.

Jin Wu, Zhili Sun

Implementation and Evaluation of DSMIPv6 for MIPL

Abstract

Mobile IPv6 performs mobility management to mobile node within IPv6 networks. Since IPv6 is not widely deployed yet, it’s an important feature to let Mobile IPv6 support mobile node moving to IPv4 network and maintaining the established communications. DSMIPv6 specification extends Mobile IPv6 capabilities to allow dual stack mobile nodes to move within IPv4 and IPv6 networks. This paper describes the implementation of this feature based on open source MIPL under Linux. By performing experiments on testbed using the implementation, it is confirmed that the DSMIPv6 works as expected.

Mingli Wang, Bo Hu, Shanzhi Chen, Qinxue Sun

Grid Management: Data Model Definition for Trouble Ticket Normalization

Abstract

Handling multiple sets of trouble tickets (TTs) originating from different participants in today’s GRID interconnected network environments poses a series of challenges for the involved institutions. Each of the participants follows different procedures for handling trouble incidents in its domain, according to the local technical and linguistic profile. The TT systems of the participants collect, represent and disseminate TT information in different formats. As a result, management of the daily workload by a central Network Operations Centre (NOC) is a challenge on its own. Normalization of TTs to a common format for presentation and storing at the central NOC is mandatory.

In the present work we provide a model for automating the collection and normalization of the TT received by multiple networks forming the Grid. Each of the participants is using its home TT system within its domain for handling trouble incidents, whereas the central NOC is gathering the tickets in the normalized format for repository and handling. Our approach is using XML as the common representation language. The model was adopted and used as part of the SA2 activity of the EGEE-II project.

Dimitris Zisiadis, Spyros Kopsidas, Matina Tsavli, Leandros Tassiulas, Leonidas Georgiadis, Chrysostomos Tziouvaras, Fotis Karayannis

Extension of Resource Management in SIP

Abstract

In this work we discuss the issue of communication QoS management in a high performance network oriented to carry GRID application. Based on a set of previous works that successfully proved the feasibility of the concept, we propose to use sessions to logically identify and mange the communication between the applications. As a consequence the quality of service of the communication is mapped on reservation of network resources to support a given session. Starting from a framework defined to support such a task for VoIP applications we show here how this framework can be extended to match the need of GRID computing.

Franco Callegati, Aldo Campi

Economic Model for Consistency Management of Replicas in Data Grids with OptorSim Simulator

Abstract

Data Grids are currently solutions suggested meeting the needs for the scale large systems. They provide whole of resources varied, geographically distributed of which the goal is to ensure a fast access and effective to the data, to improve the availability, and to tolerate the breakdowns. In such systems, these advantages are not possible that by the use of the replication. Nevertheless, several problems can appear with the use of replication techniques, most significant is the maintaining consistency of modified data. The strategies of replication and scheduling of the jobs were tested by simulation. Several simulators of grids were born. One of the important simulators for Data Grids is the OptorSim tool. In this work, we present an extension of OptorSim by consistency management service of replicas in data Grid. This service is based on hybrid approach, combining between pessimistic, optimistic approaches and economic model, articulated on a hierarchical model.

Ghalem Belalem

A Video Broadcast Architecture with Server Placement Programming

Abstract

We propose a hybrid architecture MTreeTV to support fast channel switching. MTreeTV combines the use of P2P networks with dedicated streaming servers, and was proposed to build on the advantages of both P2P and CDN paradigms. We study the placement of the servers with constraints on the client to server paths and evaluate the effect of the server parameters. Through analysis and simulation, we show that MTreeTV supports fast channel switching (<4s).

Lei He, Xiangjie Ma, Weili Zhang, Yunfei Guo, Wenbo Liu

VXDL: Virtual Resources and Interconnection Networks Description Language

Abstract

Data grid applications require often an access to infrastructures with high performance data movement facilities coordinated with computational resources. Other applications need interconnections of large scale instruments with HPC platforms. In these context, dynamic provisioning of customized computing and networking infrastructure as well as resource virtualization are appealing technologies. Therefore new models and tools must be studied and developed to allow users create and handle such on-demand virtual infrastructures within grid platforms or even within the Internet. This work presents VXDL, a language for virtual resources interconnection networks specification and modeling. Besides allowing end resources description, VXDL lets users describe the desirable virtual network topology, including virtual routers and timeline. In this paper we motivate and present the key features of our modeling language. We explore typical examples to demonstrates the expressiveness and the pertinence of it. Then we detail experimental results based on the execution of NAS benchmark on virtual infrastructures, conforming different VXDL specifications.

Guilherme Piegas Koslovski, Pascale Vicat-Blanc Primet, Andrea Schwertner Charão

Adding Node Absence Dynamics to Data Replication Strategies for Unreliable Wireless Grid Environments

Abstract

To the existing data replication approaches, we add more dynamics about occasional data absence whereby files are replicated in advance in order to make all sites as suitable for job executions as possible. Occasional node or edge absence resulting in data absence is inevitable in wireless network environments. This approach offers several advantages to wireless grids, such as improved fault-tolerance, and less overhead due to dynamic data replication.

Soomi Yang

Hop Optimization and Relay Node Selection in Multi-hop Wireless Ad-Hoc Networks

Abstract

In this paper we propose an efficient approach to determine the optimal hops for multi-hop ad hoc wireless networks. Based on the assumption that nodes use successive interference cancellation (SIC) and maximal ratio combining (MRC) to deal with mutual interference and to utilize all the received signal energy, we show that the signal-to-interference-plus-noise ratio (SINR) of a node is determined only by the nodes before it, not the nodes after it, along a packet forwarding path. Based on this observation, we propose an iterative procedure to select the relay nodes and to calculate the path SINR as well as capacity of an arbitrary multi-hop packet forwarding path. The complexity of the algorithm is extremely low, and scaling well with network size. The algorithm is applicable in arbitrarily large networks. Its performance is demonstrated as desirable by simulations. The algorithm can be helpful in analyzing the performance of multi-hop wireless networks.

Xiaohua(Edward) Li

Localization Anomaly Detection in Wireless Sensor Networks for Non-flat Terrains

Abstract

In this paper, we develop and evaluate a Localization Anomaly Detection (LAD) scheme for non-flat surfaces for wireless sensor networks. The beacon-less grid localization scheme proposed for non-flat terrains in [1] is used for localization and the localization anomaly detection uses observations of the sensor node at two different reception ranges. Moreover, a new Signal Strength (SS) Metric is proposed and evaluated for LAD. Simulations show that the beacon-less localization method when combined with the LAD scheme gives good detection rates with low false positive rates for the proposed Signal Strength Metric and the Difference Metric. The results show that Signal Strength Metric is comparable to existing metrics while being more difficult to attack.

Sireesha Krupadanam, Huirong Fu

Network Coding Opportunities for Wireless Grids Formed by Mobile Devices

Abstract

Wireless grids have potential in sharing communication, computa-tional and storage resources making these networks more powerful, more robust, and less cost intensive. However, to enjoy the benefits of cooperative resource sharing, a number of issues should be addressed and the cost of the wireless link should be taken into account. We focus on the question how nodes can efficiently communicate and distribute data in a wireless grid. We show the potential of a network coding approach when nodes have the possibility to combine packets thus increasing the amount of information per transmission. Our implementation demonstrates the feasibility of network coding for wireless grids formed by mobile devices.

Karsten Fyhn Nielsen, Tatiana K. Madsen, Frank H. P. Fitzek

Automatic Network Services Aligned with Grid Application Requirements in CARRIOCAS Project

(Invited Paper)

Abstract

Automatic Service framework named Scheduling, Reconfiguration and Virtualization (SRV) is developed in CARRIOCAS project to enhance existing Telecom network infrastructures for supporting grid applications sharing IT resources interconnected with ultra-high performance optical networks. From the requirements of Grid applications a classification is pro po sed to specify the network services and their attributes. In large-scale collaborative environments, SRV solution is described to enable automatic network service operations according to high-performance computing service access. The resources hosted at datacenters are virtualized to be attached to transport network infrastructure offering uniform interfaces towards external cus tomers. New level of service bindings is defined with network services du ring executions of Grid applications’ workflows. On-demand intensive com puting and visualization services scenario is described in Telecom environment.

D. Verchere, O. Audouin, B. Berde, A. Chiosi, R. Douville, H. Pouyllau, P. Primet, M. Pasin, S. Soudan, T. Marcot, V. Piperaud, R. Theillaud, D. Hong, D. Barth, C. Cadéré, V. Reinhart, J. Tomasik

Communication Contention Reduction in Joint Scheduling for Optical Grid Computing

(Invited Paper)

Abstract

Optical network, which can provide guaranteed quality of service (QoS) connections, is considered as a promising infrastructure for grid computing to solve more and more complex scientific problems. When optical links are regarded as resources and jointly scheduled with other grid resources, communication contention must be taken into consideration for efficient task scheduling. This paper models the optical grid computing as a communication-aware Directed Acyclic Graph (DAG) scheduling problem. To reduce the communication contention, we propose to use hop-bytes metric (HBM) heuristic to select computing resource. Simulation results show that the HBM approach combined with the adaptive routing scheme can achieve better performance in terms of normalized schedule length and link utilization.

Yaohui Jin, Yan Wang, Wei Guo, Weiqiang Sun, Weisheng Hu

Experimental Demonstration of a Self-organized Architecture for Emerging Grid Computing Applications on OBS Testbed

Abstract

As Grid computing continues to gain popularity in the industry and research community, it also attracts more attention from the customer level. The large number of users and high frequency of job requests in the consumer market make it challenging. Clearly, all the current Client/Server(C/S)-based architecture will become unfeasible for supporting large-scale Grid applications due to its poor scalability and poor fault-tolerance. In this paper, based on our previous works [1, 2], a novel self-organized architecture to realize a highly scalable and flexible platform for Grids is proposed. Experimental results show that this architecture is suitable and efficient for consumer-oriented Grids.

Lei Liu, Xiaobin Hong, Jian Wu, Jintong Lin

Joint Scheduling of Tasks and Communication in WDM Optical Networks for Supporting Grid Computing

Abstract

A flexible task model (FTM) is proposed for modeling the relationship between grid tasks. We investigate the problem of scheduling grid computing tasks under FTM using light-trails in WDM networks to supporting the data communication between the tasks. Simulation results show that our proposed task scheduling algorithm under FTM significantly reduces the total task completion time.

Xubin Luo, Bin Wang

Manycast Service in Optical Burst/Packet Switched (OBS/OPS) Networks (Invited Paper)

Abstract

Recently there is an emergence of many Internet applications such as distributed interactive simulations (DIS), and high-performance scientific computations such Grid computing. These applications require huge amount of bandwidth and a viable communication paradigm to coordinate with multiple sources and destinations. In this work we propose variation of multicasting called quorumcasting or manycasting. In manycasting destinations are to be determined rather than given unlike in the case of multicasting. We first present a need to support manycasting over OBS networks. Quality of Service (QoS) policies implemented in IP does apply does not apply for optical burst switched (OBS) networks, as the optical counterpart for store-and forward model does not exist. Hence there is a need to support QoS for manycasting over OBS networks. In this work we focus on QoS parameters such as contention, optical signal quality, reliability, and propagation delay. Burst loss in OBS network can occur due to contention or bit-error rate (BER). We propose algorithms to decrease the overall burst loss. We show that IP based manycasting has poor performance compared to our proposed algorithms. Our simulation results are verified with the help of analytical model. This work is further extended as to multi-constrained manycast problem (MCMP). In this problem, we address the burst scheduling for multiple QoS constraints. We propose algorithms to minimize burst loss based on given service requirements. The goal of this work is to develop service-oriented optical networks (SOON) for many emerging Internet applications.

Vinod M. Vokkarane, Balagangadhar G. Bathula

OBS/GMPLS Interworking Network with Scalable Resource Discovery for Global Grid Computing

(Invited Paper)

Abstract

In recent years, Grid computing is more common in the industry and research community and will open to the consumer market in the future. The final objective is the achievement of global Grid computing, which means that the computing and networks are flexibly integrated across the world and a scalable resource discovery scheme is implemented. In this paper, a promising architecture, i.e., optical burst switching (OBS)/generalized multi-protocol label switching (GMPLS) interworking network with Peer-to-Peer (P2P)-based scheme for resource discovery is investigated to realize a highly scalable and flexible platform for Grids. Experimental results show that this architecture is suitable and efficient for future global Grid computing.

J. Wu, L. Liu, X. B. Hong, J. T. Lin

Providing QoS for Anycasting over Optical Burst Switched Grid Networks

Abstract

This paper presents a mathematical framework to provide Quality of Service (QoS) for Grid Applications over optical networks. These QoS parameters include, resource availability, reliability, propagation delay, and quality of transmission (QoT). These multiple services are needed to ensure the successful completion of a Grid job. With the help of link-state information available at each Network Element (NE), the bursts are scheduled to its next link. This de-centralized way of routing helps to provide optimal QoS and hence decreases the loss of Grid jobs due to multiple constraints.

Balagangadhar G. Bathula, Jaafar M. H. Elmirghani

A Paradigm for Reconfigurable Processing on Grid

Abstract

In a Grid environment, it is hard and even impossible to perform scheduler performance evaluation in a controllable manner as resources and users are distributed across multiple organizations with their own policies. To overcome this limitation a Java-based discrete-event grid simulation toolkit has been developed that called GrdiSim.

Mahmood Ahmadi, Stephan Wong

Erratum to: Networks for Grid Applications

Abstract

Erratum to: P. Primet et al. (Eds.)Networks for Grid Applications DOI: 10.​1007/​978-3-642-02080-3

The book was inadvertently published with an incorrect name of the copyright holder. The name of the copyright holder for this book is: © ICST Institute for Computer Science, Social Informatics and Telecommunications Engineering. The book has been updated with the changes.

Pascale Vicat-Blanc Primet, Tomohiro Kudoh, Joe Mambretti

Backmatter