Press Archive

SDSC SRB Version 3.3.1 Manages Data for Large Scientific Collaborations

SDSC Storage Resource Broker Meets Need for Production Data Management

Published 05/03/2005

By Paul Tooby
SDSC Senior Science Writer

When a large eScience collaboration in the United Kingdom needed to provide comprehensive data management for 16 separate sites connected to a central data repository, the SDSC Storage Resource Broker, known as the SRB, was one of the few systems they selected for an evaluation that is still underway. With innovative features designed to support reliable performance in large-scale scientific collaborations, version 3.3.1 of the SRB is a key component of emerging cyberinfrastructure -- sophisticated technology-supported environments for advancing scientific knowledge -- and the most proven data management software available for the demanding requirements of such major projects. The SRB version 3.3.1 software, user manuals, and release notes with a complete list of new features are online for download by the research community as a source distribution at www.sdsc.edu/srb/indx.php/Main_Page.

"The advanced capabilities of the SRB have made it a reliable production tool for scientific data management," said Reagan Moore, Distinguished Scientist and Director of the Data-Intensive Computing Environments group at the San Diego Supercomputer Center. "The SRB system is used in scientific disciplines from astronomy and environmental sciences to neurosciences, physics, and chemistry." The broad adoption of the SRB software is also shown by its use in projects with a large number of agencies, including the National Archives and Records Administration, the National Science Foundation, the National Institutes of Health, the Department of Energy, and the National Aeronautics and Space Administration, as well as UK eScience projects and other international efforts.

As an end-to-end data management solution, the SRB can store data created in sensor networks or simulations, support data management and collaboration in data grids, publication in digital libraries, and long-term preservation in digital archives. The flexible Zone SRB system can manage data from simple collections for a single researcher to complex multi-terabyte collections (one terabyte is 1,000 gigabytes) with millions of files. By supporting the federation of distributed data collections, Zone SRB allows scientific users to flexibly share rapidly-changing data collections across multiple institutions that may be spread around the globe, each running its own SRB Metadata Catalog (MCAT), or zone. This gives researchers powerful data discovery tools as well as speedy access to "any data, anywhere, anytime."

In conjunction with version 3.3.1 of the SRB, the SDSC team is releasing updated versions of inQ, the Windows graphical user interface to SRB, MySRB, the web-based access tool, and Jargon, a Java application program interface.

What's New in SRB Versions 3.3 and 3.3.1
As a widely-used production tool, the evolution of the SRB is guided by input from its large user community. "We spend a lot of time supporting the people who use our software," explained Arcot Rajasekar, who leads the SRB project at SDSC. "Not only does this help our users, it also helps the team better understand what users need most. We are also now using the Bugzilla system to track bugs, and recently completed a user survey that gave us good ideas for new features."

An important capability requested by many SRB users, including the eScience project of the Biotechnology and Biological Sciences Research Council, is support for transparent data movement and archiving in large research environments that have complex architectures. The UK researchers need to store their data at each local site initially, where scientists can work with it efficiently. Then, when a data collection is no longer being used locally, behind the scenes the administrator needs to be able to move the data to the central repository. After a time in the central cache, it may be moved to archival storage, but the data always needs to be easily available to move back to local storage when a scientist needs to work with it.

To support this capability, the SRB team implemented high performance capabilities that allow the administrator to easily move data collections between local storage and the central repository. The features include bulk move for shifting even millions of files, and parallel transfers for rapidly moving terabyte-size data collections. For users, all of this is transparent and without changes to file and path names, so that they can focus on their science and easily access their data whenever they need it, without needing to know technical details such as the data's physical location, access protocols, or authentication steps.

Along with replication for data backup, an important benefit of this capability is that it enables projects to build increasingly large data collections. "This is a useful model for many collaborations because it lets them flexibly scale up to very large data collections simply by providing a large central repository, while continuing to rely on relatively modest local storage resources," said Rajaseker.

Version 3.3 also includes support for GridFTP in Globus Grid Toolkit® 3 as another way for users to transfer data into and out of their SRB collections. GridFTP is a secure data transfer protocol optimized for the high-bandwidth wide-area networks of grid computing. The SRB has for some time supported the related Grid Security Infrastructure, known as GSI, and SRB version 3.3.1 now supports the latest version of this security protocol. An additional security feature allows SRB users to scramble passwords.

An important new capability makes it simpler for users to set up access control by allowing newly created collections to be easily assigned the same access control permissions that a parent collection has. Simply by setting a software control for a collection, any new SRB object or collection created, copied, or moved into this collection will inherit all of the access control permissions from the parent collection.

One of the core capabilities of the SRB is its powerful support for metadata, or information about the data in a collection, which makes it possible for users to create, manage, search, and share extremely large and complex data collections with millions of files and tens of terabytes of data. At the request of the Southern California Earthquake Center TeraShake collaboration, which is simulating large earthquakes in Southern California, the SRB team has made a major effort to add extensible schema metadata, that is, more sophisticated forms of descriptive information about the collection. Previously, users could add any number of single metadata attributes, for example, location for earthquake data, along with the corresponding value, which let them flexibly annotate their data. With the new capability, users can now include tables (which have both rows and columns) in their metadata, extending the SRB from linear to 2-D metadata.

Environmental science is another field that is driving the addition of new capabilities to the SRB. Environmental researchers are taking advantage of new sensor technologies to gather a wide range of data about the world, often as real-time data streams. But this growing volume of sensor data poses a major challenge for scientists trying to gather, manage, and analyze this information. To overcome this problem, the SRB team is collaborating with the Real-time Observatories, Applications, and Data management Network project, or ROADNet, which collects seismic, oceanographic, and other types of environmental data in Southern California. The SRB team has extended the SRB to handle two real-time data resources used in many seismic sensor networks. Built by Boulder Real Time Technologies, Inc., the first is an object ring buffer system called the Antelope Real Time System, which provides real-time acquisition, transport, buffering, processing, archiving, and distribution of environmental monitoring information. The second resource is a companion database called Datascope, designed for real-time data. By supporting access to these resources, the new version of the SRB lets scientists gather packet-based data streams from multiple sensor networks, advancing environmental science.

SDSC SRB is supported on a wide variety of systems. The MCAT Metadata Catalog runs on Oracle, IBM DB2, Sybase, Informix, MySQL, and Postgres databases. The SRB Server runs on Microsoft Windows NT, 2000,and XP, as well as most UNIX platforms, including Linux, IRIX, AIX, HP Tru64, and Mac OSX. The SRB Server supports data in file systems, tape stores such as the High Performance Storage System (HPSS), and databases. In addition to UNIX clients, additional application program interfaces include C and C++ library calls, Shell commands, Perl and Python load libraries, dynamic load libraries for Windows, Open Archives Interface, WSDL, and Java classes. Interactive browser interfaces include the Windows graphical user interface, inQ, and the Web interface, MySRB.

The SRB team, led by Reagan Moore and Arcot Rajasekar, includes chief architect Michael Wan, senior developer Wayne Schroeder, data grid application specialist George Kremenek, SRB administrator Sheau-Yen Chen, and data grid developers Charles Cowart, Lucas Gilbert, Arun Jagatheesan, Sifang Lu, Roman Olschanowsky, Antoine de Torcy, Tim Warnock, and Bing Zhu. -Paul Tooby