Session: CERN

David Foster of CERN
A Brief Tour of Grid Computing for the LHC.
Particle Accelerator

1/8th of the tunnel was already the largest superconductor ever.
1.9 Kelvin
15 PetaBytes of new data each year (that's what's stored, but it produces a PB/second)

Tier 1: (CERN) data acquisition, initial processing, and distribution
Tier 2: (Distributed World-wide) data heavy analysis. Managed mass storage.
Tier 3: Simulation and End-user analysis - batch and interactive

Any Tier 2 may access data at any Tier 1.

Middleware: Security, data management, job management, information system

EGI: European Grid Initiative. www.eu-egi.org

The accelerator is constructed, but still needs to be proved. Two sections have been cooled, with the remaining six to go. CERN is the largest consumer of liquid hydrogen in Europe. Each magnet is 10 meters in length and weighs 30 tons, they expand/contract so the connections are tricky. After multiple cooling/warming cycles, the fingers connecting the magnets got bent, causing a multi-year delay. The plan is to cool it once and leave it cold.

The target to open is this year. April 6th is the last day for the public to go to the tunnel because after that it will be radioactive. There's a press event later in the year (September?). The reality is they will be delighted if it ever works.

Cost so far: 10-15 billion Euro.

Can run about 100 days a year, the rest of the time it's in maintenance. Have robots to go into the radioactive tunnels and do work. Although the radiation dissipates to a point where people can go in on occasion.

The energy in the bean is equivalent to 100kg of TNT. And since it's in a very small space, it can drill a hole through anything imaginable. So the energy of the beam needs to be distributed over the target.

600 million proton collisions per second.

100 meters underground.

We had time at the end, so a discussion regarding the beginnings of the Internet was sparked. The funny part? It was originally for academic and research, if you put anything commercial on there, "You were shot."

Session: NorStore

NorStore - A National Distributed Storage Infrastructure for Research & Education
By Jan Meijer

Norway is a mountainous country, rocks. Hard to dig to lay cable. 4 High Performance Computing Locations.

Why? separate long term storage from HPC. Can lose information when changing computers. Need a strategy, policy, and practice regarding the creation, management, and long term storage of data.

Trends: Size of data has increased to terabyte scale. IT tools evolve rapidly and the flexibility in using these tools put the very data they create and transform at risk. Survival of digital scientific information depends on a hierarchy of constantly shifting technologies.

Reasons to keep data:
- retention of unique observational information which is impossible to recreate
- retention of expensively generated data which is cheaper to mainteain than to recreate.
- reuse data for new or future research purpose
- validate and account for publicly funded research
- for compliance with legal requirements
- for educational and teacher purposes.

Objective: establish and maintain a broad and sustainable infrastructure for the curation, archiving,... [?]

Goals:
-operate storage resources and peripheral equipment
-provide support to researchers needing storage capacity, digital repositories, and curation services.
-promote a set of standard services and establish best practices and polices that aim to improve the reuse and reusability of scientific data
-provide easy, secure, and transparent access to distributes storages resources, provide access to larger aggregate storage.

Non-technical issues: security, confidentiality and continued privacy, ownership, assured provenance, authenticity and integrity. How to guarantee the quality of the primary data and associated metadata? One solution to some of this is to only handle data that has open-access, so ownership, etc goes away.

Initial consortium: Universities in Bergen, Trondheim (NTNU), Oslo, Tromso, UNINETT (NREN), UNINETT Sigma

2007: initial specs, choice of technologies, levels of curation, roadmap for 2008, and accumulation of experience
2008: investment in hardware for initial infrastructure.

Initially two locations with 600TB each.

http://www.norstore.no
http://www.terena.org/activities/storage