Release Notes 2 1
This document describes the new features of SRB 2.1, released on June 3, 2003. For additional information, see the SRB home page at http://www.sdsc.edu/srb. Also see ReleaseNotes2.0 for more information on the SRB 2.0.0 (released 2/18/2003), and intermediate releases 2.0.1, and 2.0.2.
The GridFTP driver, which we had planned to include in this release, is being postponed to the next sub-release (probably a few weeks). We have seen problems when transferring more than a few files, which seem to be related to the GridFTP servers.
The other planned features listed on our future plans page at http://www.sdsc.edu/srb/srbFuture.html are included in this release, and are items 1, 2, and 3 below. The additional items were more recently included for this release.
A few bugs have been fixed, the more significant of which are described off of our bugs page at: http://www.sdsc.edu/srb/bugs.html
1) Better support for GSI (Grid Security Infrastructure). This was actually released early, as part of 2.0.2, but is included here as it was listed on the srbFuture.html page for 2.1. It is now easier to build SRB with support for GSI, via some new configure options. The separate AID library is no longer needed as the SRB release contains the needed source code that interfaces to the GSI libraries. We had expected that other changes would be needed to interface to GSI 2.2 (NMI 2), but found that none were.
2) Optional data encryption and compression. We had planned to implement GSI data encryption on network traffic, but found we could make it more secure, higher-performance, and also include data compression via a different approach. This system is accessed via scripts that operate on top of SRB Scommands and OpenSSL commands. Some of the SRB commands have been enhanced to encrypt key fields (file encryption keys) as they are transferred to/from the MCAT-enabled SRB server. This SRB field encryption is performed via OpenSSL libraries, which are the also the foundation of GSI. For more information, see README.dataEncryptionCompression and http://www.sdsc.edu/srb/SecureAndOrCompressedData.html.
3) SDSC Matrix, a service (SOAP/WSDL) oriented architecture of SRB. SDSC Matrix is a W3C standards based approach to deliver multi-organizational Information Storage Management Services, which can be used to provide a (web) service oriented architecture for datagrids, digital libraries and persistent archives. These include services for data movement, replication, access control, data set ingestion, retrieval and, container support. SDSC Matrix it is a protocol layer on top of the SRB protocol stack providing value-addition in terms of interoperability, asynchronous messages and stateful services. Matrix internally converts the SOAP protocol based messages to native SRB protocol messages. This is the first of a series of releases for SDSC Matrix, as it is under active development. See http://www.sdsc.edu/srb/matrix for more information.
4) JARGON, a pure Java API. JARGON is a pure java API for developing programs with a grid interface. The API currently handles file I/O for local and SRB file systems and is easily extensible to other file systems. File handling with JARGON closely matches file handling in Sun's java.io API, a familiar API to most java programmers. The previous SRB Java client library (used by SrbBrowser and the admin tool) was implemented via JNI, which connects to the C client code. See http://www.sdsc.edu/srb/jargon for more information.
5) Bulk load without requiring the use of a container. The Sbload utility has been modified to allow the bulk load of the content of a local directory to SRB without requiring the use of containers. If the [-c container] option is not specified in the command line, the bulk load is assumed to be carried out without the use of a container. In addition, a new API - srbBulkLoad() has been created for such operations. The performance improvement without the use of a container is not as great as a bulk load using a container since the container design fits the bulk load operation perfectly. Nevertheless, improvement up to a factor of 5 can be achieved. The Sbunload of files that are not in containers will be next.
6) Listing of host-specific resources. A [-H hostname] option has been added to the SgetR utility to list all resources on the specified host.
7) Parallel transfer configuration. Two configurable parameters, MAX_THREAD and SIZE_PER_THREAD, were added to determine the number of threads for parallel transfer. These can be adjusted in the runsrb script, before starting a server.
8) SRB Python binding. The new SRB Python binding exposes some SRB client APIs which allow any Python program to access SRB systems via POSIX style file I/O calls, to query SRB system metadata for files, and to manipulate SRB systems such as creating and deleting a collection. A file, README.python, included in SRB source tree, provides an overview of SRB Python binding and an example of using SRB Python binding APIs.
9) Sput/Sget checksum, status, and retry features added. Sput -M will store the checksum of the original file into the MCAT as it is being stored. Sget -M will verify this checksum on the data as the file is retrieved. Both of these operate within the Sput/Sget programs, without additional disk I/O. Sget or Sput -V displays a progress bar. -v shows a final message. -V forces -v. Sput or Sget -R number, will cause it to retry up to 'number' times if an error occurs, with an exponential backoff sleep in between.
10) SRB_Install_Notes.doc - This is an excellent SRB installation note written by Dr. Michael Doherty of the e-Science Group at the Rutherford Appleton Laboratory of UK. It describes from start to finish the installation of MCAT, MES (MCAT enable server), other SRB resource servers and client softwares. The install note is based on a Linux system with Oracle 9i as the database for MCAT. However, most of the note is applicable to other OS and DMBS.