Agenda - BNL, 18-22 October, 2004

Abstract:

This talk presents the current status, activities and plans of the INFN TRIP Project. The aim of the project is the definition and development of an authorization and authentication infrastructure which allow any authorized user to access to wireless network services of any INFN section. At this time we have implemented an architecture based on Mac address authentication, X.509 certificate or username and password to access the service.

This system use captive portal, dinamic VLAN assignment and MAC address database (via Radius protocol).

We are evaluating a solution that integrates the system with 802.1x authentication.

Abstract:

This talk will cover current work on LCG/EGEE security policy and procedures, including changes for LHC User Registration

Abstract:

We describe our work on GUMS, a site tool for Resource Authorization (AuthZ) and Grid User Identity Mapping. We will first define the scope of the work, and describe the general direction we are taking. We will describe the current functionalities provided to BNL, such as the ability to have a flexible site policy controlled by a single XML file and the ability to integrate with site databases. We will then describe the current work being done for OSG, which includes the use of account pool, a GT3/4 based service and role based authorization using VOMS extended proxy credentials.

SATA based LINUX file server can be used as relatively cheap disk storage. Important questions are the reliability and the performamance in RAID setups. This talk presents our experimences and perfomance tests taking into account different (kernel and RAID) configurations.

Abstract:

The RHIC Computing Facility at BNL has performed an evaluation of Panasas, a distributed, networked storage solution. The objective of the evaluation was to find a fast, scalable, fault tolerant, and easy to manage product that offered a global namespace. The discussion will include a review of our current NFS/SAN architecture, Panasas, testing methodologies, expectations and results.

Abstract:

We present a condor-based system for managing data reconstruction jobs in RHIC experiments.

Abstract:

As part of the LCG Application Area the persistency framework (POOL) has been developed which has just passed the first large scale deployment in experiment data challenges resulting in some 400TB of stored data. As an extension to POOL this year a Relation Abstraction Layer (RAL) has been introduced which will couple database back-end services like Oracle and MySQL as storage for meta data.

To prepare the database deployment side in LCG the 3D project has recently been started to define together with the experiments and service providers at Tier 1 and 2 the service infrastructure hosting POOL and other relational data. This project will also propose a data distribution environment which will connect the data base services provided as part of LCG.

The CERN database group is preparing for the deployment of databases for physics by consolidating the currently many disconnected services to a small number of Linux based Oracle clusters which are expected to run RAC. I will give a brief overview of these three activities focussing on their impact in the basic computing infrastructure.

Abstract:

I will present a new Perl module which provides APIs to the following AFS monitoring and debugging tools:
afsmonitor, cmdebug, rxdebug, scout, udebug, xstat_cm_test, xstat_fs_test
I will present implementation details, usage and some real life example applications for the Nagios monitoring system.

Abstract:

In this talk I will review very briefly the goals of the quattor toolsuite and how it works, then I will present the current experience in the use of quattor in two different sites: why they use quattor, the advantages over other fabric management tools, and the problems found so far.

Abstract

Having described our new farm management tools in last HEPiX Meetings, ELFms with Leaf, Lemon, and Quattor, I will concentrate in this talk on how we handle day-to-day work with these tools, on our various farms in the CERN Computer Centre. This includes software/kernel updates, hardware exchage/failures, and how we cope with the increasing demand for more machines/clusters managed by quattor for other groups at CERN.

Abstract:

This talk presents the design of a highly available linux file service based on Open Source tools. Two identical computers in a master/slave shared-nothing setup provide a hot-standby solution. The package heartbeat is used for the communication between the nodes and for IP address takeover. File system synchronisation between both nodes is done using DRBD which mirrors a whole block device via network (a kind of network raid-1). Additionally, service reliability and server status are monitored using the tool mon. Tests show that ongoing operations are not aborted during failover and recovery is completely transparent to the clients.

Abstract:

The CERN Windows Terminal Server Service was started on 1st April 2004. It provides to the CERN community an useful and easy way to:

• run Windows applications independently of the platform where you are
• securely access the CERN environment from outside CERN.
• provide a service that can be used as base to setup customized clones for special needs (e.g. controls).

We will present the service, its architecture, its functional components, how to profit from it and our experience along these first six months of running the service.

Abstract:

The new distribution list service is now being deployed at CERN. It is integrated on the actual Mail infrastructure. It benefits from it on several aspect: anti-spam, anti-virus, flood control, less spam sent by the system to the owners, ... A new web interface developed on ASP.NET technology is available which introduces new features: bulk operations, managing archives, ...

Abstract:

I'll briefly introduce these new platforms, showing that there's more to them than just an extended address space. Then the results of performance comparisons for physics codes on several systems with Opteron, Nocona, Prescott and conventional x86 CPUs will be shown. I'll also talk about experiences with managing 64-bit Linux on these systems.

Abstract:

The requirements of computing power increase considerably in many applications. Often it is impossible to expand the datacenter due to limitations of time and money. Using smaller computers like 1 U servers or blade systems is a possibility to build more computing power into the available building.

At the same time the exectric power consumption, and the heat dissipation, of a single computer system increases extremely. This results in a very large heat load concentrated into very small enclosures. The problem is to remove the heat and bring it outside.

In order to overcome these difficulties the Grid Computing Centre Karlsrune (GridKa) determined to install water cooled computer cabinets. Since 2 years ago the first cabinets work without any problem.

Abstract

DESY is one of the world-wide leading centers for research with particle accelerators and a center for research with synchrotron light. The hadron-electron collider HERA houses four experiments which are taking data and will be operated until 2007.

The H1 and ZEUS collaborations at HERA-II have started to exploit the Grid for their growing demand for Monte Carlo events after the update of the collider.

DESY participates in a number of national and international Grid projects. Among them are the German e-science initiative D-GRID, the EU-project Enabling Grid for E-sciencE (EGEE), and the International Lattice Data Grid (ILDG). In addition activities have started to exchange simulation data within the International Linear Collider Community (ILC).

The Grid Infrastructure at DESY is based on the so-called DESY Production Grid which make use of the LCG-2 middleware and is operated in the LCG TestZone. It is set up to provide generic Grid services for all Grid activities at DESY. Therefore the DESY Production Grid incorporates all necessary elements of a complete and independent Grid. This includes a Resource Broker, catalog services for replica and meta data management, and a dCache-based Storage Element which provides access to the entire DESY data space of 0.5 PB to the Grid.

In the contribution to HEPiX we will describe the DESY Production Grid in the context of the DESY Grid activities and present operation experiences.

 

Contributions from FNAL, CERN, SLAC and BNL plus open discussion.

Abstract:

A description of the tools developed by the Grid Operation Center to monitor LCG.

Abstract:

This talk will present plans for a new Security Operations group, including incident response and service challenges

Markus Schulz and Ian Bird

Introduction and strawman ops model for discussion