C. Meirosu

The base-line DataFlow system of the ATLAS trigger and DAQ

The base-line design and implementation of the ATLAS DAQ DataFlow system is described. The main components of the DataFlow system, their interactions, bandwidths, and rates are discussed and performance measurements on a 10% scale prototype for the final ATLAS TDAQ DataFlow system are presented. This prototype is a combination of custom design components and of multithreaded software applications implemented in C++ and running in a Linux environment on commercially available PCs interconnected by a fully switched gigabit Ethernet network.

Native 10Gigabit Ethernet experiments over long distances

The current solutions for transmitting data over Wide Area Networks (WANs) are expensive and require protocol translation at layer 1. The IEEE recently standardized the 10Gigabit Ethernet (10GE) WAN PHY as a native gateway from the Local Area Networks (LAN) to the WAN. This opened a debate as to whether Ethernet is now a valid alternative to Synchronous Optical Network/Synchronous Digital Hierarchy (SONET/SDH) for WANs. In this article, we report on the experience gathered while building the first trans-European native 10Gigabit Ethernet testbed based on WAN PHY. We describe and analyze network tests with a 1700km Ethernet network. Our work validates this approach and indicates that Ethernet can offer a large bandwidth to long-distance bulk data transfers at a trans-European level.

Performance of the final Event Builder for the ATLAS Experiment

Event data from proton-proton collisions at the LHC will be selected by the ATLAS experiment in a three level trigger system, which reduces the initial bunch crossing rate of 40 MHz at its first two trigger levels (LVL1+LVL2) to ~3 kHz. At this rate the Event-Builder collects the data from all read-out system PCs (ROSs) and provides fully assembled events to the the event-filter (EF), which is the third level trigger, to achieve a further rate reduction to ~ 200 Hz for permanent storage. The event-builder is based on a farm of O(100) PCs, interconnected via gigabit Ethernet to O(150) ROSs. These PCs run Linux and multi-threaded software applications implemented in C++. All the ROSs and one third of the event-builder PCs are already installed and commissioned. We report on performance tests on this initial system, which show promising results to reach the final data throughput required for the ATLAS experiment.

Testing and modeling Ethernet switches and networks for use in ATLAS high-level triggers

The ATLAS second level trigger will use a multilayered LAN network to transfer 5 Gbyte/s detector data from /spl sim/1500 buffers to a few hundred processors. A model of the network has been constructed to evaluate its performance. A key component of the network model is a model of an individual switch, reproducing the behavior measured in real devices. A small number of measurable parameters are used to model a variety of commercial Ethernet switches. Using parameters measured on real devices, the impact on the overall network performance is modeled. In the Atlas context, both 100 Mbit and Gigabit Ethernet links are required. A system is described which is capable of characterizing the behavior of commercial switches with the required number of nodes under traffic conditions resembling those to be encountered in the Atlas experiment. Fast Ethernet traffic is provided by a high density, custom built tester based on FPGAs, programmed in Handel-C and VHDL, while the Gigabit Ethernet traffic is generated using Alteon NICs with custom firmware. The system is currently being deployed with 32 100 Mbit ports and 16 Gigabit ports, and will be expanded to /spl sim/256 nodes of 100 Mbit and /spl sim/50 GBE nodes.

Transatlantic Native 10 Gigabit Ethernet Experiments: Connecting Geneva to Ottawa

Local area network technology has evolved rapidly in speed and distance coverage. In this paper, we report on the first transatlantic native ”local area network” built using 10 Gigabit Ethernet technology. The article introduces 10 Gigabit Ethernet as defined by the IEEE 802.3ae standard and describes tests carried out with the 10 GE WAN PHY in the laboratory and in the field. Measurements performed using traffic generators at wire speed, and results of data transfers using established and experimental protocols, are presented. We make the case for the WAN PHY as a reliable technology to enable future GRID applications over legacy long haul networks.

Guest editorial: Network and service virtualization

The introduction of software-defined networking (SDN) and network function virtualization (NFV) has altered, in a wholesale manner, the way to plan for network infrastructure evolution in the forthcoming decade. This Feature Topic aims to provide a concise reference entry point to a wider audience with respect to carrier-grade networking and service virtualization with emphasis on automating the entire networking and cloud infrastructure. The first steps in network virtualization are already taking place today, although more often than not through a piecemeal approach that simply replaces hardware-based network appliances with software-based alternatives. This may help to reduce operator costs in the mid-term but does not alter the full life cycle of service creation and deployment. The real potential for network and service virtualization lies in upgrading the entire toolbox network operators have at their disposal, as state-of-the-art research and development efforts already indicate. For example, the European FP7 UNIFY project defines an architecture where the entire network, from home devices to data centers, forms a unified production environment, a dynamic service creation platform able to distribute functions and state anywhere in the network, aided by automated orchestration engines; see www.fp7-unify.eu for more details. While drafting the Call for Papers for this Feature Topic, our first goal was to attract high-quality contributions from operator and industry research labs as this topic is particularly pertinent to practitioners in the field. Carriers, in particular, can be the main beneficiaries from the emerging infrastructures based on NFV, SDN, and cloud technologies. Therefore, articles by authors working at global operators currently developing, evaluating, and standardizing solutions for network and service virtualization were particularly welcome and encouraged. In this sense, we are glad that all five selected papers for publication in this issue are penned by experts affiliated with European, American, and Asian carriers.

Self-Management for a Network of Information

Network management has been traditionally considered in the last stages of the design and development cycles of network technology. The typical modus operandi relies mostly on networkers that manually configure active elements, troubleshoot problems as they arise and reconfigure equipment accordingly. In fact, we note that several initiatives in the Future Internet research area do not take proactive steps towards developing self-managing network architec- tures. This position paper presents the first steps taken towards adopting self-management from the initial design stages of NetInf, an information-centric, clean-slate architecture for the Future Internet, proposed by the 4WARD project. We introduce a preliminary set of guidelines on how to apply in-network management principles to Ne- tInf. Through scenarios, we examine the advantages of tightly inte- grating self-management functionality and network awareness with actual service delivery in NetInf.

The ATLAS Event Builder

Event data from proton-proton collisions at the LHC will be selected by the ATLAS experiment in a three-level trigger system, which, at its first two trigger levels (LVL1+LVL2), reduces the initial bunch crossing rate of 40 MHz to ~3 kHz. At this rate, the Event Builder collects the data from the readout system PCs (ROSs) and provides fully assembled events to the Event Filter (EF). The EF is the third trigger level and its aim is to achieve a further rate reduction to ~200 Hz on the permanent storage. The Event Builder is based on a farm of 0(100) PCs, interconnected via a Gigabit Ethernet to 0(150) ROSs. These PCs run Linux and multi-threaded software applications implemented in C++. All the ROSs, and substantial fractions of the Event Builder and Event Filter PCs have been installed and commissioned. We report on performance tests on this initial system, which is capable of going beyond the required data rates and bandwidths for Event Building for the ATLAS experiment.

Performance of the Final Event Builder for the ATLAS Experiment

Event data from proton-proton collisions at the LHC will be selected by the ATLAS experiment by a three level trigger system, which reduces the initial bunch crossing rate of 40 MHz at its first two trigger levels (LVL1+LVL2) to ~3 kHz. At this rate the Event-Builder collects the data from all Read-Out system PCs (ROSs) and provides fully assembled events to the the Event-Filter (EF), which is the third level trigger, to achieve a further rate reduction to ~200 Hz for permanent storage. The Event-Builder is based on a farm of 0 (100) PCs, interconnected via Gigabit Ethernet to 0 (150) ROSs. These PCs run Linux and multi-threaded software applications implemented in C++. All the ROSs and one third of the Event-Builder PCs are already installed and commissioned. Performance measurements have been exercised on this initial system, which show promising results that the required final data rates and bandwidth for the ATLAS event builder are in reach.

Investigation of the networking performance of remote real-time computing farms for ATLAS trigger DAQ

To test the feasibility of using remote farms to perform real-time event selection in the Trigger/Data Acquisition System for the ATLAS experiment at CERN, a Proof of Concept was set up during 2004. The behavior of the request-response protocol to move application data has been measured for remote farms connected with different Wide Area Networks including a dedicated lightpath, a Virtual Private Network, and the standard production network. The dynamics and effect of using TCP/IP as the transport protocol has also been investigated for this real-time application. These data are compared with conventional bulk data transfers and used to validate the observed performance of the online farms and estimate the effect of this traffic pattern on the Wide Area Network