Panayiotis Tsanakas

On Quality of Service Support for Grid Computing

Computing Grids are hardware and software infrastructures that support secure sharing and concurrent access to distributed services by a large number of competing users from different virtual organizations. Concurrency can easily lead to overload and resource shortcomings in large-scale Grid infrastructures, as today’s Grids do not offer differentiated services. We propose a framework for supporting quality of service guarantees via both reservation and discovery of best-effort services based on the matchmaking of application requirements and quality of service performance profiles of the candidate services. We illustrate the middleware components needed to support both strict and loose guarantees and the performance assessment techniques for the discovery of suitable services.

Parallel approaches to piecewise linear approximation

Abstract Two parallel algorithms have been developed for the piecewise linear approximation (PLA) of digitised curves. The first one is a new general purpose PLA algorithm, based on certain improvements of a serial algorithm. The second one is a peak preserving PLA algorithm particularly suited for the ECG waveform approximation. Both algorithms have been fully implemented, tested and evaluated on a distributed memory parallel architecture, using the OCCAM language. The derived results for both algorithms are encouraging, since they lead to optimal curve approximations, and they are amenable to real-time PLA applications.

A grid middleware for data management exploiting peer-to-peer techniques

In this paper, we describe a service-oriented middleware architecture for Grid environments which enables efficient data management. Our design introduces concepts from Peer-to-Peer computing in order to provide a scalable and reliable infrastructure for storage, search and retrieval of annotated content. To ensure fast file lookups in the distributed repositories, our system incorporates a multidimensional indexing scheme which serves the need for supporting both exact match and range queries over a group of metadata attributes. Finally, file transfers are conducted using GridTorrent, a grid-enabled, Peer-to-Peer mechanism that performs efficient data transfers by enabling cooperation among participating nodes and balances the cost of file transfer among them. The proposed architecture is the middleware component used by the GREDIA project, in which both media and banking partners plan to share large loads of annotated content.

Automated integration of wireless biosignal collection devices for patient-centred decision-making in point-of-care systems.

The proper acquisition of biosignals data from various biosensor devices and their remote accessibility are still issues that prevent the wide adoption of point-of-care systems in the routine of monitoring chronic patients. This Letter presents an advanced framework for enabling patient monitoring that utilises a cloud computing infrastructure for data management and analysis. The framework introduces also a local mechanism for uniform biosignals collection from wearables and biosignal sensors, and decision support modules, in order to enable prompt and essential decisions. A prototype smartphone application and the related cloud modules have been implemented for demonstrating the value of the proposed framework. Initial results regarding the performance of the system and the effectiveness in data management and decision-making have been quite encouraging.

Fall detection and activity identification using wearable and hand-held devices

Human motion data captured from wearable devices such as smart watches can be utilized for activity recognition and emergency event detection, especially in the case of elderly or disabled people living independently in their homes. The output of such sensors is streams of physical activity data that require real-time recognition, especially in emergency situations. This paper presents a novel application that utilizes the low-cost Pebble Smart Watch together with an Android device (i.e. a smart phone) and allows the efficient capturing, transmission, storage and processing of such motion data. The paper includes the technical details of the stream data capturing and processing methodology, along with a comparison of the major algorithms used for the classification of physical activity type (i.e. Mild, Moderate, Intense and Sleep). An initial evaluation of the achieved accuracy in recognizing activity type, calculating the energy consumption and detecting falls, is also included and the corresponding results are discussed. The reported results are quite promising and can enable the development of intelligent systems, capable of analyzing human behavior and triggering alarms related to human activity in addition to fall detection.

Hyperplane Grouping and Pipelined Schedules: How to Execute Tiled Loops Fast on Clusters of SMPs

This paper proposes a novel approach for the parallel execution of tiled Iteration Spaces onto a cluster of SMP PC nodes. Each SMP node has multiple CPUs and a single memory mapped PCI-SCI Network Interface Card. We apply a hyperplane-based grouping transformation to the tiled space, so as to group together independent neighboring tiles and assign them to the same SMP node. In this way, intranode (intragroup) communication is annihilated. Groups are atomically executed inside each node. Nodes exchange data between successive group computations. We schedule groups much more efficiently by exploiting the inherent overlapping between communication and computation phases among successive atomic group executions. The applied non-blocking schedule resembles a pipelined datapath, where group computation phases are overlapped with communication ones, instead of being interleaved with them. Our experimental results illustrate that the proposed method outperforms previous approaches involving blocking communication or conventional grouping schemes.

A Versatile Architecture for Building IoT Quantified-Self Applications

The abundance of activity trackers and biosignal sensors as well as the evolution of IoT and communication technologies have considerably advanced the concept of Quantified-Self. Nowadays there are several frameworks and applications that realize the concept, focusing though strictly on specific areas, from daily use to professional activities such as sport and healthcare. This work proposes a versatile, cross-domain solution for building quantified-self applications exploiting the capacities for open-design, modularity and extensibility of the AGILE IoT gateway.

Computing frequent itemsets in parallel using partial support trees

A key process in association rules mining, which has attracted a lot of interest during the last decade, is the discovery of frequent sets of items in a database of transactions. A number of sequential algorithms have been proposed that accomplish this task. On the other hand, only few parallel algorithms have appeared in the literature. In this paper, we study the parallelization of the partial-support-tree approach Goulbourne et al. (2000). Numerical results show that this method is generally competitive, while it is particularly adequate for certain types of datasets.

A tile size selection analysis for blocked array layouts

Efficient use of the memory hierarchy is essential for good performance due to the ever-increasing gap between processor and memory speed. Program transformations such as loop tiling have been shown to be an effective approach to improving locality and cache exploitation, especially for dense matrix scientific computations. In conjunction with tiling, several experimental studies have been conducted on blocked data layouts, as a data transformation technique used to boost the cache performance. The stability of the achieved performance improvements are heavily dependent on the appropriate selection of tile sizes, taking into account the actual layout of the arrays in memory. In this paper, we first provide a theoretical analysis for the cache and TLB performance of blocked data layouts. According to this analysis, the optimal tile size that maximizes L1 cache utilization, should completely fit in the L1 cache, to avoid any interference misses. We prove that when applying optimization techniques, such as register assignment, array alignment, prefetching and loop unrolling, tile sizes equal to L1 capacity, offer better cache utilization, even for loop bodies that access more than just one array. Increased self-or/and cross-interference misses are now tolerated through prefetching. Such larger tiles also reduce lost CPU cycles due to less mispredicted branches. Results are validated through simulations and actual benchmarks on various modern platforms.

Real-time medical collaboration services over the web.

The gradual shift in modern medical practice, from working alone clinical doctors to MDTs (Multi-Disciplinary Teams), raises the need of online real-time collaboration among geographically distributed medical personnel. The paper presents a Web-based platform, featuring an efficient medical data management and exchange, for hosting real-time collaborative services. The presented work leverages state-of-the-art features of the web (technologies and APIs) to support client-side medical data processing. Moreover, to address the typical bandwidth bottleneck and known scalability issues of centralized data sharing, an indirect RPC (Remote Process Call) scheme is introduced through object synchronization over the WebRTC paradigm.