CERN: RD45
From Devwiki
CERN is the world's largest particle physics laboratory. It is also famous as the organization that spawned the World Wide Web. In fact, Tim Berners-Lee, the British architect of the WWW and his Belgian collaborator, Robert Cailliau, once sat a few doors from our first customer at CERN. The formal name of the site is "Organisation Européenne pour la Recherche Nucléaire", or "European Organization for Nuclear Research." Maybe the European bureaucrats can explain the link between the acronym and the name. :-)
The RD45 Project was tasked with identifying the database technology to be used for the Large Hadron Collider project, the largest scientific experiment planned in the 20th. Century and due to come online in the first decade of the 21st. Century. The LHC is being funded and built in collaboration with over two thousand physicists from 34 countries, universities and laboratories. Conservative estimates put the cost of the project at over US $10 Billion. It is planned to be the largest database in the world, ingesting around eight Petabytes per year.
Contents[hide] |
[edit] Objectivity Case History
FORMER CUSTOMER |
[edit] Please Note
This information is an archive, so any use of the present sense in the text should be taken in the historical context, generally determinable from the Status section below.
[edit] Customer Information
- Customer: CERN
- Project: RD45
- Location: Geneva, Switzerland
- Territory: Europe
- Industry Verticals: High Performance Computing, Scientific Computing.
- Technologies: Very Large Database
- Application Domain: High Energy Physics experimental data processing.
- Market Characterization: Relationship Hunters, Design and Simulation
- Number of developer licenses: 10?
- Runtime license volume and type: 100? We were attempting to arrange a project based license.
[edit] Status
- First Contact: 1994
- Lead came from: Objectivity ex-employee at Supercomputing Conference
- Evaluation Start Date: 1995
- Evaluation Finish Date: 1996
- First Purchase Date: 1996
- Deployment Date: Some projects deployed in 1996 and 1997, but not the LHC.
- Current Status: No longer in maintenance mode. No new projects are anticipated.
- Can we talk about this customer and the product/project? Yes, there were several interesting applications.
- Referenceable?: No. We were unable to agree the final details of the licensing terms, so the project was told to find an alternative. They initially used Oracle to index structured data files that are in a format named ROOT.
[edit] Environment
- Hardware: DEC Alpha
- Operating System: Ultrix
- Precision: 64 bit
- Development language: C++ and Java
- Compiler: DEC
- Third Party vendor tools: NAG libraries, TGS and other visualization software
- Open Source tools: Linux (Trial use only)
[edit] The Project
[edit] Project Background
The customer said "When we started the RD45 project, we were focussing on the needs of the experiments that will take data at the Large Hadron Collider (LHC), which is indeed not yet operational. It will enter operation in around 2005."
It will actually enter production in 2007 using Oracle.
[edit] Project Description
Note: This was written in early 1998
Objectivity/DB is being used by existing experiments using existing, operational accelerators at CERN, including the Super Proton Synchrotron (SPS), and Large Electron Positron collider (LEP). Experiments at the SPS that are using Objectivity/DB in production include NA45, NA48 and CHORUS. Experiments at LEP are testing Objectivity/DB, but not yet using it in production. COMPASS, an experiment on the SPS will start in 1999 to take approximately 400TB of data per year. In December 1998, they will test writing at 35MB/s into Objy.
The LHC experiments, even though the accelerator is not yet built, are already using Objy for production work. This includes storing the results of detector simulation programs and taking data from test detectors at other accelerators at CERN. Experiments at the LHC include ATLAS, CMS, ALICE and LHCb.
Many federations have been built. The largest production federations are still around 100GB, although ATLAS should have 600GB by end-August (maybe earlier) as part of a test beam activity. In a similar test beam last year, CMS took 100GB of data in about 1000 databases.
We do not intend to backup these federations. Most of the data will reside on tape in HPSS managed storage, and will be cached to disk when required. Some parts of the federation, including the catalog & schema, will require backup. Also data that is "expensive" to recreate, e.g. took many 1000s of CPU hours to generate. We do not yet have a backup strategy, but are developing backup procedures.
[edit] Related Projects
- CERN: AMS
- CERN: ATLAS
- CERN was instrumental in persuading the SLAC: BaBar project to use Objectivity/DB.
- TRW contacted CERN for a reference during their evaluation process and were impressed by their level of commitment (at that time).
[edit] Product Influence
- CERN accelerated (sic) the development of Objectivity/DB Very Large Database features, including OOFS, the HPSS interface and Java based DBA tools (using Eclipse).
- CERN tested as many boundary conditions (number of databases, number of containers per database etc.) as possible and verified that attempting to exceed the limits was handled gracefully. Also:
- The boundary tests verified the scalability of the Objectivity/DB address space.
- They confirmed known shortcomings in the scalability of the database catalog, which used linear searches for name matching at that time.
- They verified the architecturally anticipated need for Container Files.
- CERN successfully tested database replication over high bandwidth links between Geneva, Switzerland and Caltech in Pasadena, CA.
- CERN confirmed that low bandwidth, high latency networks posed a problem by running tests over a satellite link between clients in Tokyo, Japan and data and lock servers in Geneva, Switzerland.
[edit] Buying Criteria
[edit] Business Priorities
- Strong vendor support
- Flexible licenses
The buying criteria was the believe in 1995 that Objectivity/DB was the only commercial solution that could meet the scalability that CERN needs in 1999 and beyond. The scalability of Objectivity/DB at the moment is limited to only 6PB. Although far beyond the other solution, CERN needs to store in excess of 100PB per year. CERN believes that this is feasible given the architecture of Objectivity/DB.
[edit] Technical Priorities
The buying criteria was the belief in 1995 that Objectivity/DB was the only commercial solution that could meet the scalability that CERN needs in 1999 and beyond. The scalability of Objectivity/DB at the time was limited to only 6PB. The customer said "Although far beyond the other solution, CERN needs to store in excess of 100PB per year. CERN believes that this is feasible given the architecture of Objectivity/DB."
Note: We extended the address space to millions of Petabytes later by adding the Container File feature.
[edit] Competitors/Alternatives
Oracle, Versant, ObjectStore and O2
[edit] Why They Chose Objectivity
CERN built a comprehensive benchmark and we met or exceeded all requirements. We also promised to port to the High Performance Storage System (HPSS).
[edit] Partners
DEC was a strong and well respected partner at CERN.
[edit] Collateral
- Press Releases: None
- Flyers: None
- White Papers:
- Other:
- A Google search for Objectivity and CERN will find many related pages.]
[edit] Contact Information
[edit] Important Message
Do not call this ex-customer.
[edit] Contact Details
- Objectivity Representative: General Manager, Europe
- Customer:
- Contact: Was Jamie Shiers
- Customer Phone: Do not contact
- Customer Email: Do not contact
- URL: http://wwwinfo.cern.ch/asd/rd45/index.html
[edit] Related Pages
- Application/Project List
- Customer List
- Former Customers
- Industry Verticals
- Market Characterizations
- Technologies
