Daniele Zito

A P2P strategy for QoS discovery and SLA negotiation in Grid environment

In recent years, Grid systems and Peer to Peer networks are the most commonly-used solutions to achieve the same goal: the sharing of resources and services in heterogeneous, dynamic, distributed environments. Many studies have proposed hybrid approaches that try to conjugate the advantages of the two models. This paper proposes an architecture that integrates the P2P interaction model in Grid environments, so as to build an open cooperative model wherein Grid entities are composed in a decentralized way. In particular, this paper focuses on a QoS aware discovery algorithm for P2P Grid systems, analyzing protocol and explaining techniques used to improve its performance.

Service Virtualization Using a Non-von Neumann Parallel, Distributed, and Scalable Computing Model

This paper describes a prototype implementing a high degree of transaction resilience in distributed software systems using a non-von Neumann computing model exploiting parallelism in computing nodes. The prototype incorporates fault, configuration, accounting, performance, and security (FCAPS) management using a signaling network overlay and allows the dynamic control of a set of distributed computing elements in a network. Each node is a computing entity endowed with self-management and signaling capabilities to collaborate with similar nodes in a network. The separation of parallel computing and management channels allows the end-to-end transaction management of computing tasks (provided by the autonomous distributed computing elements) to be implemented as network-level FCAPS management. While the new computing model is operating system agnostic, a Linux, Apache, MySQL, PHP/Perl/Python (LAMP) based services architecture is implemented in a prototype to demonstrate end-to-end transaction management with auto-scaling, self-repair, dynamic performance management and distributed transaction security assurance. The implementation is made possible by a non-von Neumann middleware library providing Linux process management through multi-threaded parallel execution of self-management and signaling abstractions. We did not use Hypervisors, Virtual machines, or layers of complex virtualization management systems in implementing this prototype.

Cognitive Application Area Networks: A New Paradigm for Distributed Computing and Intelligent Service Orchestration

Distributed systems are dynamic systems where software and hardware together deliver information processing services to allow modelling, interaction, reasoning, analysis and control of the external environment. The intent of the distributed computation is to execute computational workflows using computing resources. The software contains the algorithms that specify the tasks while the hardware provides the required resources to execute the algorithms. The initial structure is defined by the association of software with hardware and the dynamic structure is defined by their temporal evolution. The meta-knowledge of the intent of the algorithm, the association of specific algorithm to a specific device, their temporal evolution and exception handling when the computation deviates from the intent is outside the software and hardware design and is expressed in non-functional requirements. In this paper, we describe an architecture to capture the meta-knowledge in meta-containers and enforce the intent of the computation while the computation is in progress.

Scalable and Configurable Monitoring System for Cloud Environments

The scale of the current cloud infrastructures and the heterogeneity of applications running on them pose significant challenges about resources management on clouds. This paper proposes a monitoring system that allows user to collect and aggregate data in a form flexible and adequate to the integration with the management policy that he adopt. This monitoring system is supplied by a a configurable and highly scalable version of publisher-subscriber pattern. It gives user the capability of selecting data he should obtain from the provider with the desired grain of triggering and is suitable to support distributed and cooperative management systems.

Detecting Attacks on Java Cards by Fingerprinting Applets

Differently from classical JVMs, generally JavaCard Virtual Machines (JCVMs) rely only on limited amount of resources typical of smart cards. Recently, several mixes of logical and physical manoeuvres have been devised to exploit typical JCVM weaknesses and to have mutant applets by means e.g. of faults injection. Such combined attacks manage to bypass the existing countermeasures of several versions of JCVMs. This paper proposes an approach for detecting mutant applets, as soon as they appear at runtime, hence revealing and thwarting an ongoing attack. The approach is completely transparent to the applets developer, and only a limited (a priori computable) amount of resources are used at runtime, hence it is affordable on board of smart cards. Our experiments have shown that the proposed solution has only a very limited impact on the performances of the target platform.

Advanced Reservation in Grid Using PBS

Provisioning specific level of QoS in a grid environment represents a central concern in obtaining a flexible and dynamic services management on large scale distributed systems. The recent orientation of grid middlewares towards the concept of trading services, wherein providers and consumers of services are able to negotiate all services parameters, confirms this trend. The introduction of mechanisms to handle the advance reservation in grid, allows the system administrator to manage the resources in an optimal way and guarantees to users about the possibility to obtain a specific resources for a specific time range. This paper proposes a model of advance reservation, integrated within an more general architecture able to manage SLA in a grid environment, based on a priority mechanism for the management of the allocation request. The developed reservation component has been integrated into OpenPBS scheduler.

Evaluating the Performance of A4SDN on Various Network Topologies

Load balancing is one of the main goals of Software-Defined Networking (SDN). Distributing the traffic among the possible paths of a network according to specific rules allows SDN systems to increase their global performance index in term of resource usage and scalability. In a previous paper, the authors introduced Adaptive Alienated Ant Algorithm for Software-Defined Networking (A4SDN), a distributed, adaptive, load-balancing algorithm for traffic engineering on SDN showing its ability to optimise the network performance in terms of throughput, communication delay and packet loss rate. In this paper, the authors analyse how the performance of A4SDN are influenced by the underlying topology. For each considered topology, the performance of A4SDN are successfully with those of Extended Dijkstra algorithm. The tests show that the performance of A4SDN are the best when the average number of available shortest paths for each couple of nodes rises.

UCMS: User-side cloud management system

IaaS cloud providers, as Amazon or Rackspace, provide their users with the abilities to create an own virtual environment, populate it with their applications, and manage a wide range of critical issues (e.g. scale up/down, load balancing, faults) by means of ad hoc services. One of the limits of these cloud systems is the lack of tools that enable the user to have a direct knowledge of the physical resources supporting his application. Some providers supply a set of tools for the monitoring of the user application (e.g. CloudWatch) but these present several limits: the monitoring time is constrained by the provider, the mechanism to get the values of the state is not always clear, the observation procedure introduces an overhead unquantifiable and, in some cases, not always negligible. To overcome such a limitation, this paper introduces UCMS, an autonomic user-side management system to monitor and control the behaviour of users applications on a Cloud environment. The main advantages offered by UCMS is its ability to modify at run-time the Components-Virtual Machines-Physical Machines mapping, previously established by both the user (components-virtual machines) and cloud provider (virtual machines-physical machines) in order to improve the exploitation of the assigned resources. Experiments led on real scenario (Amazon EC2) demonstrate the ability of UCMS to reduce the execution time of considered cloud applications (Pipeline and MapReduce) by about 50%.

QoS Aware Services Discovery in a P2P Grid Environment

In recent years, grid systems and peer to peer networks are the most used solutions to achieve the same goal: sharing resources and services in heterogeneous, dynamic, distributed environments. Grid systems have well defined middleware, able to assure trust among participants, but they are not totally scalable and they have low fault tolerance. On the other hand, with their decentralized model, P2P systems are more scalable and fault tolerant than grid systems, but they do not have concrete solutions for issues related to security or credential management. Many studies have proposed hybrid approaches that try to conjugate the advantages of the two models. This paper proposes an architecture that integrates the P2P interaction model in grid environments, so as to build an open cooperative model wherein grid entities are composed in a decentralized way. In particular, this paper focuses on a QoS aware discovery algorithm for P2P Grid systems, analyzing protocols and explaining techniques used to improve its performance.

Strategies for Reserving Resources on a Grid Environment

Ensuring that a job executes at the desired timeframe in a grid environment is possible only when reservation policies are employed. Such policies often require users to predict job needs in terms of duration and CPU capability. However the efficient use of resources can be jeopardised by the low accuracy of user predictions. This paper proposes: (i) an automatic support for detecting the job characteristics and timing requirements, and (ii) a novel admission control policy relaying on such a knowledge and some indications on credential and reputation that requests could bring. The developed reservation component has been integrated into Globus toolkit in a way that let users perform just allocation, when reservation is unnecessary or impossible