1. INTRODUCTION
1. Charge to the Committee

  USATLAS Networking Committee Charter

INTRODUCTION

This document describes the need and scope of a proposed US ATLAS committee for networking.

Computing trends in physics have made a steady progression toward smaller, more local machines, starting from mainframes and moving toward PCs. Commensurate with this physical decrease in both size and distance from the user has been a significant increase in the power and number of such machines. During the same period the network has come into being, evolving from a temporary interconnect measured in bits per second to a permanent entity, independent of computer end-nodes, with maximum speeds currently measured in 100’s of Gigabits. Both of these trends are leading toward the next logical step in computer evolution: the development of grid computing.

Computers are typically thought of as a combination of CPU, memory, long-term storage and a means to interconnect these components and the user(s).

As grid computing develops, the “computer” becomes a virtual machine composed of many CPUs whose physical bus has been replaced by the network. It is the network which will provide the interconnect between CPU and storage and user and this places significant, new requirements on the network in terms of performance guarantees of bandwidth, packet loss, jitter and latency.

Independent of grid computing there will also be the traditional role the network must perform of providing access to data (http, ftp, rcp, etc.) and services (email, telnet, etc).

US ATLAS will have significant computational challenges to prepare for. The ATLAS detector will generate over 1.5 Petabytes of raw data per year.

Significant amounts of additional data will be generated simulating the detector, transforming and analyzing the raw data and providing calibration information for the detector as a function of time.

The most critical component for delivering the end-to-end performance necessary for the data transfer requirements of the Atlas project is the network. This committee will focus on networking issues to ensure that Atlas applications can successfully deliver the performance necessary to fulfill the commitments made by the Atlas project.

Back to top

Charge to the Committee

Charge to the Committee

1. Gather Atlas Application Requirements and Assess Application Network Needs

• What is the required network bandwidth and application throughput between?:

1. Tier0 and Tier1 centers
2. Tier1 and Tier2 centers
3. Tier2 and Tier3 centers

• What types of network services are required for the ATLAS computing model (QoS/CoS, new TCP protocols, diagnostics, security, authentication, authorization, policy management, etc)?

• Which USATLAS networking needs will be provided by current/future work underway or planned?

• What expectations should US ATLAS have concerning available bandwidth as a function of time for:

1. Backbone (ESNet, Abilene, Commodity networks)
2. Campus infrastructure
3. To the desktop

• What expectations should US ATLAS have concerning available network services as a function of time for:

4. Backbone (ESNet, Abilene, Commodity networks)
5. Campus infrastructure
6. To the desktop

2. Perform an Assessment of Current and Future Technologies Necessary to Meet the Requirements

• Given the planned hardware/bandwidth deployment during the next 5 years, which services and research should US ATLAS invest in to insure the success of the computing model and the experiment?

3. Develop a Deployment Plan to Apply Technologies to Meet Atlas Application Requirements

• Once an appropriate technological response to the challenges faced by Atlas has been determined, how will these technologies be deployed and supported?

4. Assess the Impact of Deployed Technologies on Atlas Application

• Which application and network performance characteristics should be fed back to the application? What infrastructure needs to be in place to do this?

Back to top