BIRN Coordinating Center — Infrastructure Primer

The Biomedical Informatics Research Network (BIRN) is an NIH-sponsored infrastructure initiative that fosters distributed collaborations in biomedical science by utilizing information technology innovations. To enable this collaborative environment, the BIRN Coordinating Center (BIRN-CC) is designing and deploying an architecture designed around a flexible large-scale grid model. A grid is defined as a collection of network-connected resources that can be accessed, scheduled, and coordinated to solve problems. This collection of network-connected component resources is tightly integrated by an evolving layer of grid middleware technologies. While there are many models for grid computing, the BIRN-CC is building on the most widely accepted model for large-scale computing and is taking advantage of many ongoing research and development efforts in industry as well as efforts funded by the National Science Foundation and other government agencies. The term "grid" is increasingly be referred to as "cyberinfrastructure." The BIRN uses these terms interchangeably.

Grid middleware enables the assembly of geographically disparate resources into an application-specific virtual resource. The low-level grid software layer for BIRN is built upon a collection of community-accepted software systems and services distributed as part of the NSF Middleware Initiative (NMI). A key component of the NMI distribution, the Globus Toolkit supplies the essential grid services layer that includes authentication, encryption, resource management, and resource reporting. Globus provides the underlying security and job submission components that allow test bed researchers to launch long running jobs to any available computing resources on the BIRN grid. For the seamless management (including submission and scheduling) of computational jobs across the BIRN grid, GSI-enabled Condor is utilized which represents the vast resources of the BIRN Grid through a simplified single interface. Grid middleware also provides the infrastructure to interconnect independent data collections, a fundamental requirement of scientific collaboration.

Within the BIRN test beds, this need was met through the implementation of the BIRN Data Grid which:

• Provides a software "fabric" for seamlessly federating data repositories across the geographically distributed BIRN sites;
• Facilitates secure access and maximizes data input/output performance across the network; and
• Facilitates the collaborative use of domain tools and flexible processing/analysis workflows.

The Biomedical Informatics Research Network employs a 'mediator architecture' to link multiple databases together, each with a unique schema, into an accessible format or data federation. The mediator architecture is flexible, scalable, and powerful. It allows individual researchers to manage their own data in databases tailored to meet their specific needs. Those searching the BIRN system, however, will see the collection of federated databases as if it were a single database. More information on data integration is available.

The BIRN Portal 2.0 provides a comprehensive collaborative environment for biomedical scientists to form new projects with robust data management functionality. Project-specific forums, blogs, and various lists foster improved communication among project members. The collaborative features support access control by assigning user or administrator roles to members. In addition, projects, blogs, and forums can be marked public or private, depending on the requirements of the research groups.

The BIRN Portal takes full advantage of the latest grid technology to ensure secure access to the BIRN testbed, including the storage resource broker (SRB). Data management features allow users to download and upload data to the SRB as well as to launch tools that can be used in conjunction with selected data, such as ImageJ and 3D Slicer. Users can apply permissions to data hosted on the SRB along with its associated metadata. For more information about SRB, please visit the BIRN SRB Central.

BIRN researchers need to rely on an ever-increasing collection of software, hardware, and policy resources to effectively and securely share data and collaborate among each other. Logically, BIRN software components include: user identification (certification), user authorization, digital document signing, data location, data mediation, data retrieval, workflow engines and domain-specific applications. These components, once defined, need to be documented, packaged, installed, configured and updated on appropriate physical hardware located throughout the BIRN test beds and the BIRN-CC.

The BIRN-CC is tasked with defining, integrating, packaging, fielding, and updating a complete cyberinfrastructure. To effectively address and manage this expanding complexity, the BIRN-CC has formalized and expanded the process of integration, testing, deployment, and updating of an integrated software stack.. BIRN-CC issues regular releases (and deploys on the BIRN physical hardware infrastructure) a complete and integrated system. This integrated software stack includes all the above described components (i.e. grid middleware for security, data and computation, data mediation and integration, and BIRN Portal), as well as the underlying system software and the BIRN Bioinformatics packaging of scientific tools and biomedical applications that have been defined by the BIRN scientific test beds (e.g. 3D-Slicer, AFNI, FSL, BRAINS2, etc.). These biomedical applications are packaged and then delivered in the Red Hat Package Manager (RPM) format, a de facto standard packaging format. The BIRN Coordinating Center coordinates and produces these software releases in collaboration with BIRN test beds to best optimize the balance of functionality, utility, and robustness. The BIRN-CC has produced internal releases of software and has deployed four successive major upgrades of the BIRN Racks during the past four years. The major release schedule incorporates a comprehensive process for the migration from alpha through beta releases to obtain a robust production release.

In order for the BIRN-CC to efficiently deploy this integrated software stack across the entire BIRN infrastructure, the BIRN-CC has adopted the NPACI Rocks Cluster Toolkit, a set of open-source enhancements for building and managing Linux™-based clusters. While BIRN racks are not traditional HPC clusters (each node in a BIRN rack has a specialized function), Rocks has enabled the BIRN-CC to automate the loading of its software stack so that after a few configuration parameters are provided, each resource is then deployed in an automated fashion. This ensures consistent configuration across all grid sites, provides greater reliability, simplifies the diagnostic efforts to correct problems, and thus increases the availability of the BIRN resources. As the BIRN continues to grow, the Rocks Cluster Toolkit will continue to evolve to support more platforms and a greater heterogeneity of site configurations.

BIRN's technologically advanced environment - by synchronizing developments in wide area networking, distributed computing, and the uniform access to heterogeneous data sources - is helping scientists and clinicians to make new discoveries. To effectively share and collaboratively explore pooled, distributed data, BIRN's powerful cyberinfrastructure is composed of an optimally integrated set of standardized hardware, software, and networking tools for seamless interoperability with Grid resources. Further information on BIRN hardware configurations are available.