200 likes | 297 Vues
This project supports various scientific communities with diverse requirements for data retrieval, management, and storage. It offers dual access via Grid and non-Grid interfaces to analyze, curate, and archive very large volumes of data with high real-time rates. The system ensures data recovery in emergencies, using tape and mass storage technologies. With a capacity of 1.2 Petabytes, STK Powderhorn tape, and multiple interfaces and services, including SRB and SRM, this project caters to a wide range of clientele from GridPP to e-Science projects and CCLRC facilities. By aligning with community needs and offering robust storage solutions, this project fosters collaboration and supports scientific breakthroughs.
E N D
Supporting Multiple Interfacesto Mass Storage Providing Tape and Mass Storage to Diverse Scientific Communities
Requirements Retrieve data Data Management Locate data
Requirements Dual access – Grid and non-Grid Analyse data on the Grid
Requirements Curation Long term storage and archival
Requirements Very large volumes Very large real-time data rates
Requirements Recover data in emergency = backup
What is Atlas? • 1.2 Petabyte capacity • Tape • STK Powderhorn • Interfaces, services
What is Atlas? • 1.2 Petabyte capacity • Tape • STK Powderhorn • Interfaces, services • New robot coming in
We support – Grid protocols • SRB • Data management interface • Metadata, data • SRM • Built for very large data volumes • Very high transfer rates via GridFTP
What is the Grid? Distributed computing Distributed collaborations and Virtual Organisations
What is the Grid? Access is brokered job Data is replicated
What is the Grid? Well defined protocols (sort of) • File access • Information providers • Job submission • Security
Grid Architecture, SRB Scientist SRB Atlas
Grid Architecture, SRB Slow network Fast network Scientist Local SRB Remote SRB Atlas Group files into container Store container
Grid Architecture, SRM Application Information Services File Transfer Service Replica Manager Storage Element (SRM) Replica Catalogue Components fit together to provide Grid services Atlas
We support – non-Grid protocols Tape Disk cache vtp, rfio, dcap,…
Who are the customers GridPP Tier 1 Research councils e-Science projects CCLRC facilities
Who are the customers Small – a few gigabytes To large – hundreds of Terabytes Different customers drive different areas of service Be all things to all people?
Community How to tie the community together? User group meetings Helpdesk
Conclusions • Supporting multiple communities via multiple interfaces • Grid interfaces and non-Grid • Multiple requirements • Diversity is good – (up to a point?) • Volume and rates driven by GridPP • Metadata driven by e-Science projects and RCUK