Odej Kao

Scheduling in HPC resource management systems: Queuing vs. planning

Nearly all existing HPC systems are operated by resource management systems based on the queuing approach. With the increasing acceptance of grid middleware like Globus, new requirements for the underlying local resource management systems arise. Features like advanced reservation or quality of service are needed to implement high level functions like co-allocation. However it is difficult to realize these features with a resource management system based on the queuing concept since it considers only the present resource usage.

Towards ontology-driven P2P grid resource discovery

In this paper we present a new approach to semantic resource discovery in the grid. A peer-to-peer network is used to distribute and query the resource catalogue. Each peer can provide resource descriptions and background knowledge, and each peer can query the network for existing resources. We do not require a central ontology for resource description and matching. Each peer has its own, possibly incomplete, ontology, which is completed by the knowledge distributed over the network. This allows to find matching resources even if the concepts used to describe the resources are unknown to the provider, as the network supplies the missing parts of the ontology.

Design of data-intensive Web-based information services

One common usage of the WWW is to provide information services, e.g. for regional event calenders, university lecture management, travel information, etc. Their design requires a clear picture of the systems usage and the supported technical environment. The core of the problem is to provide flexible access units based on extended views on some underlying database. The extension consists of escort information, links, operational facilities and means for the generation of different presentations. We call such units media objects. In order to also allow a flexible presentation for them, we adapt the metaphor of a container. In this way it is possible to present the same information in various ways depending on the user profile and the possible technical restrictions of the display device.

On triple dissemination, forward-chaining, and load balancing in DHT based RDF stores

The Resource Description Framework provides a powerful model for structured knowledge representation that allows the inference of new knowledge. Because of the anticipated scope of semantic information available in the future, centralized databases will become incapable of handling the load. Peer-to-Peer based distributed databases offer better scalability and integration of many different data sources. In this paper we present a detailed data management strategy for a DHT based RDF store that provides reasoning, robustness, and load-balancing.

Retrieval of remotely sensed imagery using spectral information content

The retrieval of images in professional remote sensing databases is based on world-oriented information like for example the location of the scene, the utilised scanner, and the date of acquisition. However, these descriptions are not meaningful for many users who have a limited knowledge about remote sensing but nevertheless have to work with satellite imagery. Therefore the retrieval of an appropriate scene is often almost impossible since adequate techniques are not provided by the database systems. This paper proposes the content-based retrieval for satellite imagery using the query-by-example approach. The main contribution is the reliable and scanner independent technique for the feature extraction and representation. The entire approach is unsupervised to stay abreast of the daily increasing huge data volume.

On the impact of reservations from the grid on planning-based resource management

Advance Reservations are an important concept to support QoS and Workflow Scheduling in Grid environments. However, the impact of reservations from the Grid on the performance of local schedulers is not yet known. Using discrete event simulations we evaluate the impact of reservations on planning-based resource management of standard batch jobs. Our simulations are based on a real trace from the parallel workload archive. By introducing a new option for scheduling reservations in planning-based resource management, less reservation requests are rejected. Our results are important for increasing the acceptability of the Grid technology. We show, that a limited number of additional resource reservations from the Grid have only a limited impact on the performance of the traditionally submitted batch jobs.

Scheduling aspects for image retrieval in cluster-based image databases

Systems for the archival and retrieval of images are used in many areas, for example medical applications, news agencies, etc. The state-of-the-art approach for image description considers a priori extracted features. The disadvantageous reduction of the image content onto a few low-level features limits the applicability of image databases. A search for objects and other important image components requires dynamic feature extraction. The related computational and storage requirements exceed the possibilities of computer architectures with a single processing element. Therefore we developed a cluster platform, which supports the implementation of this novel retrieval approach in existing systems. We introduce the basic principles of image retrieval with dynamic feature extraction and a cluster platform. The main focus regards thereby the workload balancing across the cluster. For this purpose we developed a scheduling heuristic and executed performance measurements with the implemented prototype. The obtained results are discussed.

Processing complex RDF queries over P2P networks

In large-scale distributed systems, information is typically generated decentralized. However, for many applications it is desirable to have a unified view on this knowledge, allowing to reason about it and to query it without regarding the heterogeneity of the underlying systems. In this context, two main requirements have to be fulfilled. On the one hand, a retrieval system has to be semantically rich, in order to be able to cope with and mediate between different schemas, and on the other hand it has to be scalable to large numbers of information sources. The dynamic nature of information makes the problem even worse.Within this paper, we propose a solution to this problem. We describe a DHT-based peer-2-peer network storing knowledge in the form of RDF triples. The query evaluation algorithm allows to use arbitrary query patterns, and evaluates the query with respect to taxonomical reasoning. Thus the system combines expressivity and scalability. Although we describe the whole system, the focus of this paper is the query evaluation.The system is generic by nature and suitable for numerous different applications. We describe an example application stemming from the Semantic Grid.

Top k RDF query evaluation in structured P2P networks

Berners-Lees vision of the Semantic Web describes the idea of providing machine readable and processable information using key technologies such as ontologies and automated reasoning in order to create intelligent agents. The prospective amount of machine readable information available in the future will be large. Thus, heterogeneity and scalability will be central issues, rendering exhaustive searches and central storage of data infeasible. This paper presents a scalable peer-to-peer based approach to distributed querying of Semantic Web information that allows ordering of entries in result sets and limiting the size of result sets which is necessary to prevent results with millions of matches. The system relies on the graph-based W3C standard Resource Description Framework (RDF) for knowledge description. Thereby, it enables queries on large, distributed RDF graphs.

The First Step of Introducing Risk Management for Prepossessing SLAs

Using service level agreements (SLAs) is a central concept for the grid commercialization since customers require guarantees for a successful job execution. However, agreeing on SLAs is a business risk for resource providers: system failures can lead to paying tremendous penalties. Accordingly, resource providers require an estimation about the risk of agreeing on an SLA. Furthermore, risks can be published and integrated into the negotiations as an additional SLA parameter. Each offer is linked with a possible execution on a specific resource since the risk of failure depends especially on the resources stability. Therewith making reservations is the initial step for integrating risk management into the grid. This paper describes the necessary workflow and measurements for selecting a resource according to assessed risks. A global perspective is presented in order to afford user-specific configurations. Concrete examples for several measurements present one possibility for rating and the important input factors