79 research outputs found
Προσομοίωση Συστημάτων Επεξεργασίας Δοσοληψιών και Μελέτη Μεθόδων για την Ικανοποίηση Στόχων Επίδοσης
No full textThe transaction concept, based on the concept of contract law, signifies the properties of atomicity, consistency, isolation, and durability for a sequence of actions. These properties, widely known with the acronym ACID, are essential for supporting concurrent access to shared data and for failure handling, especially in distributed environments. The object of this work is the development of an integrated environment for simulation of trancaction processing systems, with the aim to study the operation and performance of such systems. The environment incorporates the TPsim simulator and a mechanism that supports experiment management, so as to automate the process of specifying simulation experiments, conducting the experimental steps, and collecting measurements. The simulator models with considerable detail the operation of a typical trancaction processing system, and enables the estimation of a variety of variables related to resource consumption and system performance. The simulation environment was used for the experimental evaluation of a series of scheduling algorithms for complex units of work, that consist of multiple transactions. Workload units of this type occur often in practice, and represent a special class of workflows. The study of this workload class has started to draw considerable research interest, and this work presents a first performance study. The proposed scheduling algorithms are oriented towards satisfying performance goals, which are specified as requirements about the average response time per workload class. # Η έννοια της δοσοληψίας (transaction), βασισμένη στην νομική έννοια της σύμβασης, υποδηλώνει τις ιδιότητες της ατομικότητας, συνέπειας, απομόνωσης, και μονιμότητας για μια σειρά πράξεων. Οι ιδιότητες αυτές, που αναφέρονται στη βιβλιογραφία με την ακροστοιχίδα ACID (atomicity, consistency, isolation, durability), είναι απαραίτητες για την υποστήριξη ταυτόχρονης προσπέλασης σε κοινόχρηστα δεδομένα και την αντιμετώπιση βλαβών, ειδικά σε κατανεμημένα περιβάλλοντα. Στα πλαίσια της εργασίας αυτής αναπτύχθηκε ένα ολοκληρωμένο περιβάλλον για την προσομοίωση συστημάτων επεξεργασίας δοσοληψιών, με σκοπό την μελέτη της δυναμικής συμπεριφοράς και επίδοσης τέτοιων συστημάτων. Το περιβάλλον περιλαμβάνει τον προσομοιωτή TPsim και ένα μηχανισμό που υποστηρίζει την διαχείριση πειραμάτων προσομοίωσης, ώστε να αυτοματοποιηθεί κατά το δυνατόν η διαδικασία της περιγραφής πειραμάτων, του συντονισμού της εκτέλεσης των πειραματικών βημάτων, και της συλλογής μετρήσεων. Ο προσομοιωτής μοντελοποιεί με σημαντική λεπτομέρεια όλα τα υποσυστήματα ενός τυπικού συστήματος επεξεργασίας δοσοληψιών, και επιτρέπει την εκτίμηση ποικιλίας μεταβλητών σχετικών με την κατανάλωση πόρων και την επίδοση του συστήματος. Το περιβάλλον προσομοίωσης χρησιμοποιήθηκε για την πειραματική αξιολόγηση μιας σειράς αλγορίθμων χρονοπρογραμματισμού για σύνθετες μονάδες φόρτου εξυπηρέτησης, που εκτελούνται ως σειρά από δοσοληψίες. Αυτού του είδους οι μονάδες φόρτου εμφανίζονται συχνά στην πράξη, και αποτελούν μια περίπτωση της κατηγορίας μονάδων φόρτου που αναφέρεται με τον όρο {\sf workflows}. Η μελέτη αυτής της κατηγορίας έχει αρχίσει να συγκεντρώνει σημαντικό ερευνητικό ενδιαφέρον και στην εργασία αυτή γίνεται μια πρώτη μελέτη επίδοσης. Οι προτεινόμενοι αλγόριθμοι χρονοπρογραμματισμού είναι προσανατολισμένοι στην ικανοποίηση στόχων επίδοσης, που διατυπώνονται ως απαιτήσεις για τον μέσο χρόνο απόκρισης ανά κλάση μονάδων φόρτου
Exploring the impact of node failures on the resource allocation for parallel jobs
No full textIncreasing the size and complexity of modern HPC systemsalso increases the probability of various types of failures. Failures maydisrupt application execution and waste valuable system resources dueto failed executions. In this work, we explore the eect of node failureson the completion times of MPI parallel jobs. We introduce a simulationenvironment that generates synthetic traces of node failures, assumingthat the times between failures for each node are independently dis-tributed, following the same distribution but with dierent parameters.To highlight the importance of failure-awareness for resource allocation,we compare two failure-oblivious resource allocation approaches withone that considers node failure probabilities before assigning a partitionto a job: a heuristic that randomly selects the partition for a job, andSlurm's linear resource allocation policy. We present results for a casestudy that assumes a 4D-torus topology and a Weibull distribution foreach node's time between failures, and considers several dierent tracesof node failures, capturing dierent failure patterns. For the synthetictraces explored, the benet is more prominent for longer jobs, up to82% depending on the trace, when compared with Slurm and a failure-oblivious heuristic. For shorter jobs, benets are noticeable for systemswith more frequent failures
Impact of Cache Coherence on the Performance of Shared-Memory based MPI Primitives: A Case Study for Broadcast on Intel Xeon Scalable Processors
No full text<p>Recent processor advances have made feasible HPC nodes with high core counts, capable of hosting tens or even, hundreds of processes. Therefore, designing MPI collective operations at the intra-node level has received significant attention over the past years. Deriving efficient algorithms for modern HPC nodes, with complex internal topologies and memory hierarchies, is challenging. Moreover, the cache coherency protocol, and its impact on performance, further complicate algorithm design for MPI collectives. This latter concern is often only partially addressed.</p><p>In this work, we demonstrate a particularly challenging performance degradation scenario in the case of shared-memory–based MPI broadcast, on three generations of the Intel Xeon Scalable processor architecture. Based on analysis of hardware-based performance counters, we conclude that the performance degradation observed is attributed to the cache coherency protocol and the multi-socket configuration of the execution platforms examined. We present a number of novel approaches designed to amend this effect, and apply them in a cache coherency aware version of the MPI broadcast implementation. We reduce the overall latency of the broadcast operation by up to 1.5 × and 1.25 × for small and large messages, respectively.</p>
The Aurora Architecture for Developing Network-Centric Applications by Dynamic Composition of Services
No full textThis paper presents the Aurora architecture for network-centric applications. This architecture complements the CORBA and WWW=Java frameworks with support for coordination and collaboration and addresses the requirements of dynamic open environments with multiple autonomous service providers. The paper presents a detailed technical design of the run-time infrastructure that enables on-demand composition of services. A specification language, called HERMES, provides constructs for describing events, tasks, and work sessions. Aurora is based on a container framework, which provides the basis for dynamic and adaptive composition, as well as detailed monotoring and tracking. A repository service, together with a metadata framework, enable interoperability for integration of services. Keywords: workflow management, coordination, collaboration, network-centric applications, open environments Contents 1 Introduction 5 1.1 Motivation : : : : : : : : : : : : : : : : : : : : : : : : : : : ..
Data-centric privacy protocol for intensive care grids
No full textModern e-Health systems require advanced computing and storage capabilities, leading to the adoption of technologies like the grid and giving birth to novel health grid systems. In particular, intensive care medicine uses this paradigm when facing a high flow of data coming from intensive care unit's (ICU) inpatients just like demonstrated by the ICGrid system prototyped by the University of Cyprus. Unfortunately, moving an ICU patient's data from the traditionally isolated hospital's computing facilities to data grids via public networks (i.e., the Internet) makes it imperative to establish an integral and standardized security solution to avoid common attacks on the data and metadata being managed. Particular emphasis must be put on the patient's personal data, the protection of which is required by legislations in many countries of the European Union and the world in general. In this paper, we extend our previous research with the following contributions: 1) a mandatory access control model to protect patient's metadata; 2) a major security revision to our previously proposed privacy protocol by contributing with a "quality of security" quantitative metric to improve fragmented data's assurance; and finally, 3) a set of early results to demonstrate that our protocol not only improves a patient personal data's security and privacy but also achieves a performance comparable with existing approaches
Study and Bridging of Peer-to-Peer File Sharing Systems
No full textAbstract—Peer-to-peer file sharing systems have become increasingly popular over the last few years, by attracting large numbers of Internet users, who share a continuously increasing volume of data. Since the launch of Napster, the first widely known peer-to-peer file sharing network, several other filesharing systems have emerged. These systems are based on various architectures, which usually reflect different sharing policies and communities. Although current peer-to-peer systems facilitate the sharing of content among their users, they do not facilitate the sharing of content among themselves. In this paper we study two of the most popular open-source peer-to-peer file sharing systems in order to understand (i) whether their traffic patterns are similar, and (ii) whether there would be any benefit in facilitating the sharing of content between the users of different peer-to-peer systems
Understanding Scalability and Performance Requirements of I/O-Intensive Applications on Future Multicore Servers
No full text- …
