René Brunner

Network-aware summarisation for resource discovery in P2P-content networks

Many application scenarios have a read-dominated behaviour of information provision, which means that there are few updates, and that users execute frequent queries for information discovery. Discovery of content in such systems could benefit from the use of summary techniques in order to facilitate the discovery process and minimise the size of the data exchanged. This is particularly applicable within Grid computing environments where nodes (representing a group of computational and storage resources), which enable jobs to reach them may need to exchange metadata with each other about their resources. We introduce network-aware summarisation algorithms for the resource discovery in P2P-content networks, which are based on Cobweb clustering. We identify how summarisation can improve the discovery process while also improving the accuracy of the discovered resource(s). Metrics based on precision-recall are used to compare the accuracy for specific types of queries generated over the summarised content.

Towards the development of a decentralized market information system: Requirements and architecture

In a market, information about its specifications and the behavior of its participants is essential for sophisticated and efficient negotiation strategies. However, there is currently no completely researched system to provide and consult an overall knowledge of economic information in distributed markets. These markets are implemented for example by grid applications and gained importance over the last few years. This paper presents the economic information requirements and a high-level architecture overview for a decentralized market information system (DMIS). The proposed system acquires economic data in a distributed environment for providing it to individual traders or other participants in a decentralized manner. First, we outline the economic information requirements which the system needs to achieve. Therefore their properties and a privacy model has to be considered. Then we propose an architecture for the system which combines technologies of distributed information aggregation system and distributed publish-subscribe models, based on a structured overlay network. The architecture has been designed to meet both the economic information requirements and that of scalability and robustness of a large-scale distributed environment. Initial measurements confirm the proof-of-concept implementation of the existing prototype.

Core Services For Grid Markets

Markets Are A Powerful Model For The Coordination Of Distributed Systems And, In Particular, In The Face Of Incomplete Information And Changing Environments. The Application Of Markets For The Resource Allocation In Grid Systems Has Recently Been Researched As An Alternative To Traditional Approaches. However, The Proper Implementation Of Sophisticated Markets Capable Of Handling Diverse Tradingmodels (Various Auctions Types, Bargaining) And Structures (Direct Negotiation, Brokering, Etc.) Requires A Set Of Supporting Services To Provide Participants A Proper Environment To Engage In Negotiations. Gridmarket Middleware (Gmm) Is A Framework That Aims To Ease The Development Of Market Based Grid Systems. In This Paper We Present Its Architecture, The Services It Provides And Describe How They Can Be Used To Implement Diverse Market Models. We Also Discuss Our Experience With The Implementation Of Prototypes For Various Core Services.

Exploring the behaviour of fine-grain management for virtual resource provisioning

The virtualization of resources in SMP machines with Xen offers automatic round-robin based coarse-grain assignment of VMs to physical processors and manual assignment via privileged commands. However, two problems can arise in Grid environments: (1) if Xens sEDF scheduler assigns the VMs, then some processors could be over or underutilized and the VMs could receive more resources than specified, and (2) manual assignment is not feasible in a dynamic environment and also requires being aware of each nodes heterogeneity. Our approach proposes an enhanced fine-grain assignment of SMPs virtualized resources for Grid environments by means of a local resource manager (LRM). We have developed a prototype which adds a partitioning layer of subsets of physical resources. Our experimental results show that our approach achieves a flexible assignment of resources. At the same time, due to the fine-grain access, a more efficient resource assignment is achieved compared to the original mechanism.

Self-adaptive approximate queries for large-scale information aggregation

Self-adaptation enables distributed software to modify its behaviour based on changes in the operating environment. In large-scale information systems for cloud computing that use hierarchical data aggregation, self-adaption may be used to respond to an approximate query, thereby reducing use of network bandwidth and retrieval time. We present a novel algorithm that uses an Analytic Hierarchical Process (AHP) in order to apply self-adaption to approximate queries based on network-awareness. The AHP-based algorithm provides a trade-off among network usage, retrieval time and the accuracy of the retrieved results. Simulations show that the number of needed messages reduces with AHP to a constant upper bound. The retrieval time reduces to a constant factor under an increasing number of nodes. Our results demonstrate that the algorithm is able to provide responses with the required accuracy, primarily by adapting the depth of the query based on the number of messages and the network conditions.

Grid4All: Open Market Places for Democratic Grids

The Grid4All project is focused on the provision of the benefits and opportunities of Grids for everyone, including small organizations such as schools, families, non-governmental organizations, or small businesses. This involves multiple relevant and related aspects despite the scale of the global system: (i) self-management of applications as they adapt to environmental changes, (ii) the complexity of developing and using applications in that situation by multiple users, (iii) and the need to organize, govern, and regulate the community. Grid4All promotes the concept of a democratic Grid, virtual organizations and self-management systems, based on decentralized overlays. Thereby, the providers offer resources and services either for a shared-interest within a virtual organization (pooling) or for an open market across virtual organizations. These two models of distributing resources that co-exist locally, contribute to achieve global regulation. We propose an architecture according to these ideas, which are inspired by real-world cases, which include a collection of data sharing and execution services, used by collaborative applications.

Assessing a distributed market infrastructure for economics-based service selection

Service selection is an important issue for market-oriented Grid infrastructures. However, few results have been published on the use and evaluation of market models in deployed prototypes, making it difficult to assess their capabilities. In this paper we study the integration of an extended version of Zero Intelligence Plus (ZIP) agents in a middleware for economics-based selection of Grid services. The advantages of these agents compared to alternatives is their fairly simple messaging protocol and negotiation strategy. By deploying the middleware on several machines and running experiments we observed that services are proportionally assigned to competing traders as should be in a fair market. Furthermore, varying the environmental conditions we show that the agents are able to respond to the varying environmental constraints by adapting their market prices.

Information and Regulation in Decentralized Marketplaces for P2P-Grids

Large scale systems such as the grid need scalable and efficient resource allocation mechanisms to fulfil the requirements of its participants and applications while the whole system is regulated to work efficiently. Economics inspired models have shown ability to handle efficiently the allocation of resources and services, scaling up well as they are decentralized. Our model considers the arbitration of decisions at the local scope and short term, the regulation of the system at global scope, and the sharing of information between global and local environments. This paper presents a scalable model and evaluates by simulation a system where global market information circulates in aggregated and scalable form, the rate of demand by participants is globally regulated by a currency mechanism, preference is regulated by a reputation mechanism, and local regulation among competing participants is resolved by auction mechanisms. The paper shows how scalable systems benefit from distributed marketplaces supporting global information flow to regulate and optimize local and global behavior.

Uncertainty Management for the Retrieval of Economic Information from Distributed Markets

The provision of real-time and highly accurate information in large-scale distributed systems is technically difficult and approximations imply uncertainty. In economic science, however, reliable information about markets, its specifications, and the behavior of its participants is essential for sophisticated and efficient negotiation strategies. There is the need for a system that provides and allows consulting an overall knowledge of economic information in distributed markets, while managing the accuracy of information for the user. This paper evaluates the influences of uncertainties for information retrieval within distributed Grid markets. It proposes an uncertainty management component for a Decentralized Market Information System (DMIS), which regulates the accuracy of information and the number of messages for the retrieval of economic data from a scalable market. First, we analyze the properties and the completeness of information in Grid markets. Therefore, we simulate Grid market specific scenarios under complete and incomplete information provision by varying the information accessibility. The results confirm the influence of the accuracy on the stability of the market. Based on these results, an optimization mechanism, which uses approximations is introduced for the retrieval of information. The approximations are controlled by the uncertainty management to find a trade-off between the amount of messages and the accuracy of information.

Elaborating a decentralized market information system

A Decentralized Market Information System (DMIS) that aggregates and provides information about markets is an important component for achieving markets in Grid and Peer-to-Peer systems. The proposed work is the development of a framework for the DMIS, which fulfils the economic provision within the main technical requirements like scalability towards nodes and data attributes and robustness against failures. The proposed work also allows obtaining results concerning the trade-off between economic benefits and technical costs. Introducing dynamic adaptive processes promises improvements in efficiency with regards to distributed queries and routing structures. This research proposal presents and discusses the research questions and challenges, the current knowledge and the research methodology proposed for the development of the DMIS framework.