Antonia Ghiselli

GridICE: a monitoring service for Grid systems

Grid systems follow a new paradigm of distributed computing that enables the coordination of resources and services that are not subject to centralized control, can dynamically join and leave virtual pools, and are assigned to users by means of an explicit assignment functionality. The monitoring of a Grid is a multi-institutional and Virtual Organization (VO)-oriented service. It must deal with the dynamics, diversity, and geographical distribution of the resources available to Virtual Organizations, and the various levels of abstraction for modeling them. This paper presents the requirements, architecture and implementation of GridICE, a monitoring service for Grid systems. The suitability of this tool in real-life scenarios is analyzed and discussed.

An Analysis of Security Services in Grid Storage Systems

With the wide-spread deployment of Data Grid installations, and rapidly increasing data volumes, storage services are becoming a critical aspect of the Grid infrastructure. Due to the distributed and shared nature of the Grid, security issues related with state of the art data storage services need to be studied thoroughly to identify potential vulnerabilities and attack vectors. In this paper, motivated by a typical use-case for Data Grid storage, we apply an extended framework for analyzing and evaluating its security from the point of view of the data and metadata, taking into consideration the security capabilities provided by both the underlying Grid infrastructure and commonly deployed Grid storage systems. For a comprehensive analysis of the latter, we identify three important elements: the players being involved, the underlying trust assumptions and the dependencies on specic security primitives. This analysis leads to the identication of a set of potential security gaps, risks, and even redundant security features found in a typical Data Grid. These results are now the starting point for our ongoing research on policies and mechanisms able to provide a fair balance between security and performance for Data Grid Storage Services.

Performance Studies of the StoRM Storage Resource Manager

High performance disk-storage solutions based on parallel file systems are becoming increasingly important to fulfill the large I/O throughput required by high-energy physics applications. Storage area networks (SAN) are commonly employed at the Large Hadron Collider data centres, and SAN-oriented parallel file systems such as GPFS and Lustre provide high scalability and availability by aggregating many data volumes served by multiple disk-servers into a single POSIX file system hierarchy. Since these file systems do not come with a storage resource manager (SRM) interface, necessary to access and manage the data volumes in a grid environment, a specific project called StoRM has been developed for providing them with the necessary SRM capabilities. In this paper we describe the deployment of a StoRM instance, configured to manage a GPFS file system. A software suite has been realized in order to perform stress tests of functionality and throughput on StoRM. We present the results of these tests.

The INFN-grid testbed

The Italian INFN-Grid Project is committed to set-up, run and manage an unprecedented nation-wide Grid infrastructure. The implementation and use of this INFN-Grid Testbed is presented and discussed. Particular care and attention are devoted to those activities, relevant for the management of the Testbed, carried out by the INFN within international Grid Projects.

Distribution of RFC 1327 mapping rules via the Internet DNS: the INFNet distributed gateway system

Abstract There is a constant increase in the traffic load between RFC822/SMTP and X.400 mail protocols, thus new and easy to use interconnection tools are required. In particular, the duplication and decentralization of gateways between the above protocols requires that they should be usable also by non-experts. The new gateways should be maintenance free. One of the major problem is RFC1327 mapping table maintenance. A mechanism for automatic distribution of mapping rules by means of Internet Domain Name System (DNS) has been implemented and tested; the positive results are presented here. In particular, we specified an Application Programming Interface (API) and developed a library which allows a gateway to interact with DNS in order to obtain mapping rules, without the need of statical tables. Using these basic tools, we developed also a new RFC 1327 gateway which does not require local maintenance. This new gateway was distributed over our INFNet backbone, resulting in an efficient and fault tolerant two protocol (X.400 and RFC 822/SMTP) E-mail distribution backbone.

Performance of 10 Gigabit Ethernet Using Commodity Hardware

In the prospect of employing 10 Gigabit Ethernet as networking technology for online systems and offline data analysis centers of High Energy Physics experiments, we performed a series of measurements on the performance of 10 Gigabit Ethernet, using the network interface cards mounted on the PCI-Express bus of commodity PCs both as transmitters and receivers. In real operating conditions, the achievable maximum transfer rate through a network link is not only limited by the capacity of the link itself, but also by that of the memory and peripheral buses and by the ability of the CPUs and of the Operating System to handle packet processing and interrupts raised by the network interface cards in due time. Besides the TCP and UDP maximum data transfer throughputs, we also measured the CPU loads of the sender/receiver processes and of the interrupt and soft-interrupt handlers as a function of the packet size, either using standard or ¿jumbo¿ Ethernet frames. In addition, we also performed the same measurements by simultaneously reading data from Fibre Channel links and forwarding them through a 10 Gigabit Ethernet link, hence emulating the behavior of a disk server in a Storage Area Network exporting data to client machines via 10 Gigabit Ethernet.

Very Large Ensemble Ocean Forecasting Experiment Using the Grid Computing Infrastructure

Istituto Nazionale di Fisica Nucleare, Bologna, Italy Submitted to Bulletin of American Meteorological Society Revised version October 12, 2007 Corresponding author address : N.Pinardi, Corso di Scienze Ambientali, Via S.Alberto 163, 49100 Ravenna Italy, e-mail: n.pinardi@sincem.unibo.it

A Comparison of Data-Access Platforms for the Computing of Large Hadron Collider Experiments

Performance, reliability and scalability in data-access are key issues in the context of the computing Grid and High Energy Physics data processing and analysis applications, in particular considering the large data size and I/O load that a Large Hadron Collider data centre has to support. In this paper we present the technical details and the results of a large scale validation and performance measurement employing different data-access platforms-namely CASTOR, dCache, GPFS and Scalla/Xrootd. The tests have been performed at the CNAF Tier-1, the central computing facility of the Italian National Institute for Nuclear Research (INFN). Our storage back-end was based on Fibre Channel disk-servers organized in a Storage Area Network, being the disk-servers connected to the computing farm via Gigabit LAN. We used 24 disk-servers, 260 TB of raw-disk space and 280 worker nodes as computing clients, able to run concurrently up to about 1100 jobs. The aim of the test was to perform sequential and random read/write accesses to the data, as well as more realistic access patterns, in order to evaluate efficiency, availability, robustness and performance of the various data-access solutions.

Standards-Based Job Management in Grid Systems

The Grid paradigm for accessing heterogeneous distributed resources proved to be extremely effective, as many organizations are relying on Grid middlewares for their computational needs. Many different middlewares exist, the result being a proliferation of self-contained, non interoperable “Grid islands”. This means that different Grids, based on different middlewares, cannot share resources, e.g. jobs submitted on one Grid cannot be forwarded for execution on another one. To address this problem, standard interfaces are being proposed for some of the important functionalities provided by most Grids, namely job submission and management, authorization and authentication, resource modeling, and others. In this paper we review some recent standards which address interoperability for three types of services: the BES/JSDL specifications for job submission and management, the SAML notation for authorization and authentication, and the GLUE specification for resource modeling. We describe how standards-enhanced Grid components can be used to create interoperable building blocks for a Grid architecture. Furthermore, we describe how existing components from the gLite middleware have been re-engineered to support BES/JSDL, GLUE and SAML. From this experience we draw some conclusions on the strengths and weaknesses of these specifications, and how they can be improved.

Monitoring the connectivity of a grid

Grid computing is a new paradigm that enables the distributed coordination of resources and services which are geographically dispersed, span multiple trust domains and are heterogeneous. Network infrastructure monitoring, while vital for activities such as service selection, exhibits inherent scalability problems: in principle, in a Grid composed of <i>n</i> resources, we need to keep record of <i>n</i>² end-to-end paths. We introduce an approach to network monitoring that takes into account scalability: a Grid is partitioned into domains, and network monitoring is limited to the measurement of domain-to-domain connectivity. However, partitions must be consistent with network performance, since we expect that an observed network performance between domains is representative of the performance between the Grid Services included into domains.