Release Notes 2 0 2
←Older revision | Newer revision→
SRB 2.0.2 (See below for 2.0.1 and 2.0)
Released on May 1, 2003
The link http://www.sdsc.edu/srb/bugs.html gives a list of bugs fixed and features added in 2.0.2. They are:
1) In case the connection to a SQL database fails, the srbMaster may not log any message and then segfault.
2) Sls -C core dump.
3) Importing large file ( > 2 Gig Bytes) with Postgres MCAT fails.
4) Sbload failure to open SRB container in some cases.
5) Sbload cause the server with Postgres MCAT to core dump.
6) "Srmcont -a" does not work for postgres MCAT.
7) Srm of a filenames with underscores '_' does not work.
8) A problem with container associated with collection when it is full.
When the container got full, a new container with the same name was created, but the collection is still associated with the old container.
9) Sls -r" does not work with Postgres MCAT.
10)Some additional checks for NULL when dealing with $HOME variable for the
11)When the path is too long and overflow the qval length of MAX_TOKEN (200).
12)Problem with using the Java Admin tool connecting to a non-mcat-enabled srb
server to create a physical resource causing the resource path to be regisatered incorrectly.
13)Fatal SRB compilation errors under recently patched versions of Redhat
Linux 8.0 or 9.0.
14)Improvement in mcatAdmin (Java Admin tool) to clarify what one should type
in certain windows.
15)Building SRB with GSI support no longer requires a separate
installation of the AID library and manual SRB GSI configuration.
16)Fatal SRB compilation errors when building on AIX with MCAT.
Released on March 14, 2003
The link http://www.sdsc.edu/srb/bugs.html gives a list of bugs fixed and features added in 2.0.1. They are:
1) Problem building with Postgres or Sybase
2) New utility Sbunload to match Sbload
3) New utility Sbregister.
4) On Linux systems, Sbload and Sbunload will hang or segfault after an error
5) he GridPort interface to the SRB would fail because the SRB Scommands
programs would die (either gracefully or segmentation fault) when faced with the situation of no $HOME variable (while looking for $HOME/.srb/.MdasEnv and Auth files).
6) Data transfers would fail on server systems with two sets of IP addresses
(i.e. an internal and an external IP address).
7) When a file is replicated and the first copy is deleted, the remaining
copies become inaccessible
Released on February 18 2003
Subject: [Announce] SRB version 2.0 is available
(brief blurb here about what SRB is/does, pointers to websites, features, supported platforms, licensing, etc. Don't forget to warn everyone that if they update their servers, they must update their clients and their MCAT database!)
The following Unix variants are supported: * Linux Redhat 7.3, 32 and 64 bits * Solaris, 32 and 64 bits * AIX * SGI * MAC - OSX
The following Microsoft Windows versions are supported: * Windows 2000
The following databases are support for MCAT: DB2 Version 5,7 Oracle 8i and 9i Sybase Postgres 7.3.2
GETTING THE SRB SOFTWARE
SRB source code is available to academic and government organizations in the United States. Limited availability exists for other countries. Contact firstname.lastname@example.org or see http://www.sdsc.edu/srb for additional information.
This document provides additional details on the new enhancements made to SRB in release 2.0.0. An overview of these can be found in README.first.htm.
1) "Server driven" parallel I/O - The "Server driven" parallel I/O scheme has been implemented throughout the SRB infrastructure. It uses the more efficient "mover" APIs and protocol to access HPSS resources. It is the preferred data transfer mechanism for data import, export, copy and replication.
The more efficient "server driven" parallel I/O scheme is replacing the "client driven" scheme used in the "srbpput" and "srbpget" utilities. "srbpput" and "srbpget" will be phased out eventually if "Sput" and "Sget" meet user's expectation.
With the "client driven" scheme, the client is the active partner which subdivides the file to be imported/exported into segments and spawns threads to handle the import/export of each segment in parallel. Each thread independently connects to the server through the normal client/server connection mechanism. So, in effect, the client is multi-threaded but the server involves multi processes.
With the "server driven" scheme, the server is the active partner. The client listens on a control socket for instructions from the servers. Upon receiving the client request, the server with the help of information from the MCAT, plans the execution of the import/export. Typically it sends data transfer instruction to the server where the import/export resource is located. The resource server then sub-divides the file to be imported into segments and spawns threads to handle the import of each segment in parallel. Each server thread independently connects to the client's control socket for data transfer. Both client and server are are multi-threaded with this scheme.
Advantages of the "server driven" scheme are:
a) The server has better access to MCAT information to plan optimum transfer strategy.
b) Data transfer is always directly between the resource server and client with no intermediate server in between.
c) It uses the more efficient "mover" APIs and protocol to access HPSS resources.
d) Both client and server are are multi-threaded which involves less overhead and consumes less resources than multi processes.
One possible problem with the "server driven" scheme is that the client machine may sit behind a security firewall that limits the ports that can be connected to the client. In this case, the firewall must allow at least a few ports to get through and the COMM_PORT_NUM_START and COMM_PORT_NUM_COUNT parameters in the mk/mk.config file can be configured to limit the range of ports used by the client. If this is not possible, the -s option should be used with Sput and Sget to force the I/O mode to serial.
The PARA_OPR flag in the mk/mk.config file can be used to switch on/off the "server driven" scheme. By default, this switch is on. For platforms such as Windows, OSX and BSD which do not support pthread, this parameter should be turned off.
By default, Sput and Sget use serial I/O while Sreplicate and Scp use the parallel I/O API for data operations. The -m option can be used with Sput and Sget to force the I/O mode to parallel.
Two new APIs - srbObjGet() and srbObjPut() have been created to support the "server driven" parallel I/O.
2) A Mass Storage System (MSS) implementation within SRB
The motivations for implementing a MSS in SRB are:
a) Cost of licensing commercial MSS such as HPSS
b) Efficiency and performance - A MSS system that is tightly integrated with the infrastructure of the SRB can take full advantage of SRB features such as the cache system, file replication and server directed parallel I/O.
c) Elimination of duplication of features - SRB and MSS such as HPSS provide some duplication of features such as disk cache, authentication system, etc.
Components of the MSS:
a) A new type of resource "compound resource" - A compound resource may be configured to contain a pool of cache resources and a tape resource. When a user creates a file using a compound resource, the object created becomes a "compound object". The actual data of a "compound object" may reside on cache or tape or both. Unlike the SRB replica, a "compound object" always appears as a single object even though there may be multiple copies of the data. It is a simple hierarchical system where data migrate automatically between cache and tape. Data is always staged on cache automatically whenever it is accessed and migrates to tape by the system admin when more cache space is needed.
b) A set of driver functions for basic tape I/O operations - a set of driver functions for basic tape I/O operations has been incorporated into the SRB server. These functions include mount, dismount, open, close, read, write, seek, etc. Currently, the driver has only been tested for 3590 tape drives. The TAPE_DRIVE flag in the mk/mk.config file can be used to switch on/off the tape drivers. By default, this switch is off.
c) A tape library server - A tape library server for the STK silo running ACSLS software has been incorporated into the SRB system. Its primary function is to schedule and perform the mounting and dismounting of tapes. It uses the same authentication system and server framework as other SRB servers. The tape library server will be a binary release only since it uses StorageTek's CSC Developer's Toolkit which requires software licensing for any source access to software written with the Toolkit.
d) A set of tape and cache management utilities.
inittape - Label a tape and register the tape with the MCAT. lstape - List the tapes and their associated meta data in MCAT. tapemeta - Modify the metadata of a tape that has been registered with MCAT. dumptape - Dump and purge files in the cache system to tape.
e) A set of tape and cache management APIs. These are all privileged calls.
srbTapelibMntCart() - Request the Tape Library server to mount a tape. srbTapelibDismntCart() Request the Tape Library server to dismount a tape. srbGetTapeCartPri() - Get the priorities for each tape type depending on the availability of tape drives. srbDumpFileList() - Dump a list of files to tape. srbStageCompObj() - Stage a compound Object from tape to cache.
f) A set of APIs that deal with compound object metadata. These are all privileged calls. Raja did the internal MCAT calls.
srbRegInternalCompObj() - register an internal compound object. srbRmIntCompObj() - unregister an internal compound objects. srbModInternalCompObj - Modify the metadata of an internal comp object. srbRmCompObj() - unregister a compound object.
3) A new command "Sbload" to imports in bulk one or more local files and/or directories into SRB space. It is similar in functionality to "Sput" but designed to greatly improve the efficiency of ingesting a large number of small files by a) registering up to several hundreds files with MCAT with a single call instead of the normal mode of registering one file at a time and b) use of separate threads for registration and data transfer.
A new API - srbBulkRegister() was created to enable the registration of several hundred files with one call.
A "Sbunload" which is the counterpart of "Sbload" will be implemented if time permits.
4) Replica synchronization - Copies of replica can modified individually and the replica of an object can be out of sync. Now, the most recently modified copy will be marked as "dirty".
A new command - "Ssyncd" and API - srbSyncData() has been created to synchronize all copies (replica) of an SRB object with the dirty copy. A "%" character to the left of the file name of a "Sls -l" output indicates the given copy is "dirty".
5) The handling of storing data objects in a logical resource consisting of multiple physical resources has been slightly modified. Previously (version 1.1.8), when a data object is created in a logical resource consisting of multiple physical resources, the data object is automatically replicated to all the physical resources in the group by default. In order to support the idea of using multiple SRB bricks (Linux PC with 1 TBytes of disk space) to create logically a single very large disk cache, only a single copy will be created in one of the resource in the group chosen randomly. If a copy is required for each physical resources, the -a option of Sput, Scp and Ssyncd should be used.
6) 64 bits addressing platform - The SRB software has been ported to the 64 bits address platform of Linux and Solaris. The software has been tested for the client and non-MCAT-enabled server configurations. Currently, since we have no license for any 64-bits DBMS, the MCAT-enabled server configuration cannot be tested.
The ADDR_64BIT flag in the mk/mk.config file can be used to switch on/off the 64 bits addressing. By default, this switch is off.
7) Enhancement to handle family of containers - When a container is full, the SRB server automatically renames the full container "myContainer" to "myContainer.NNNN" where NNNN is a random integer number. The server then creates an empty container with the original container name "myContainer" so that the client can continue to put data into the same container without worrying about it getting full. In effect, this scheme creates a family of containers without the explicit knowledge of the owner.
The -a option has been added to the "Ssyncont" and "Srmcont" commands to synchronize and remove, respectively, all containers in the family.
8) Miscellaneous Scommand enhencements :
Scp New options:
-v - Verbose mode to print out file size and transfer rate.
-a - If the target object does not exist and the target resource is a logical resource consisting of more than one resources, then make a copy in each resources. If the target object exists, force copying all replica of the target object.
Sget New options: -m - multi-threaded parallel I/O mode. -s - serial I/O mode. -v - Verbose mode to print out file size and transfer rate.
Sls New options:
-a - list metaData associated with each file and collection. -C - list the access control list (ACL) of each file and collection.
Sput New options: -m - multi-threaded parallel I/O mode. -s - serial I/O mode. -v - Verbose mode to print out file size and transfer rate. Enhenced options: -a - If the target object does not exist and the target resource is a logical resource consisting of more than one resources, then make a copy in each resources. If the target object exists, force copying all replica of the target object.
Sreplicate New options: -v - Verbose mode to print out file size and transfer rate.
Srmcont New options:
-a - Delete all containers in a container family.
Ssycont New options:
-a - Synchronize all containers in a container family.
Smeta works as advertised :-)
Sbload and Ssyncd
9. Miscellaneous inQ Enhancements:
context (right-click) menues have been enabled for right-hand windows when you double click on an icon in the lower-right hand window it opens the file as expectced. Known Bug: long pause after double-clicking icons in lower-right hand window to open the file. metadata window streamlined to be more user-friendly.
10. Miscellaneous MySRB Enhancements:
more features have been added. These include metadata copy, extraction of metadata function, different types of object support, etc.
11. MCAT Enhancements
There are several new features addaed to the MCAT module in the SRB. These include new databases which can be used to run the MCAT as well as changes that provide new capabilities for the MCAT.These new capabilities can be categorized as follows: A. Porting of MCAT to new database systems: 1. Porting of the MCAT to Sybase database. This will also enable use with SQLServer since they share API. 2. Porting of the MCAT to Postgres database. A porting to Informix will be released in the next few patches. B. Improvements to System-level Metadata Functionality 1. Access Control for metadata. 2. Softlinking of srbObjects to another collection 3. Streamlined Auditing Functionality 4. Locking and Hiding of srbObjects 5. Compound Resource and Compound Object Support 6. Bulk Ingestion of Files Support 7. Bulk Access Control Function over all Objects in Collecions 8. Delete All from Container 9. New Miscellaneaous Metadata attributes Collection Ownership and creation-time metadata. srbObject checksum, encryption-type, compression-type. srbObject Pinning, Expiration srbObject Version Number, Segment Number
C. Improvements to User-defined Metadat Functionality 1. Metadata Extraction Functionality 2. Copy Functionality for Metadata and Annotations 3. Additional Query Capability for user-defined metadata 4. Annotations for Collections 5. User-defined Metadata for Resources and Users D. Administrative Support 1. Client-side Java GUI for MCAT Administration 2. Client exposure to Administrative APIs for Admin GUI 3. New Admin Functionality for remove: users, resources, locations,domains.
We explain each of these in some detail below:
A. MCAT Ports A.1 Porting of MCAT to Sybase: The MCAT software has been ported to run its database on Sybase. The functionality is the same as that of Oracle and DB2 based MCAT systems. There were certain features which were not natively supported by Sybase such as synonyms and aliases. These were done by defining views. Because of this there might be some slowing down of the MCAT when using the Sybase MCAT. In our normal testing, we haven't found any performance degradation, but we suspect that with larger tables there might be some slowing down which is not noticed in Oracle or DB2. We haven't done much tunability experiments with Sybase MCAT. We plan to do that in the near future when required by our user base. The Sybase port was done in such a way that one can use the same port for use with SQLServer. Even though we haven't tested this feature, we believe that it should be straightforward to use the MCAT with SQLserver using the Sybase port. More details on Sybase can be found in http://www.sybase.com/home A.2 Porting of MCAT to Postgres: Due to popular demand for MCAT on a free database, we have ported the MCAT to run on Postgres. Postges is a freeware available from http://www.postgresql.com/ and runs on multiple platforms including Linux, Solaris, Win32 and xBSD. Postgres is a full-fledged SQL providing transaction-level support and is fully ACID compliance and ANSI SQL compliance. The porting was done by Matt Clark of Aerospace Corp. and was tested by the SRB group. The Postgres port is based on ODBC API and hence talks through ODBC with the Postgres database. Hence, this port can be used for other ODBC-enabled databases also. (Note: we haven't tested the port with other ODB-enabled databases). The Postgres port provides full MCAT functionality. More information about Postgres and how to download and install the database can be found at http://www.postgresql.com/.
B. Improvements to System-level Metadata Functionality Several additions and improvements have been made in the way MCAT handles core system-level metadata. Some of the functionalities have been stream-lined or improved to proved bulk operations. Bugs uncovered over the past year or so in the version 1.1.8 have been corrected. B.1. Access Control for metadata. This functionality allows owners and co-owners of srbObjects and SRB collections to control access to associated system-level and user-defined metadata that are stored in the MCAT. In pervious versions of the SRB, the user was able to access control only the collections and the srbObjects but not the associated metadata. because of this every user was able to see the collection listing (say through Sls) as well as view and query user-defined metadata. This was unacceptable to many user communities and a requirement to show metadata based on access permission was delivered to the SRB developers. This functionality is due to that requirement. The access permissions for the metadata is inherited from the access permissions of the accosiated srbObject or collection. A user will be able to see the metadata only if he/she has 'read' permission. As in earlier versions, a user can write/modify system-level metadata only if they have ownership and a user can write/modify user-defined metadata only if they have 'write' permission. The access control for metadata is made into an option which can be turned on or of at compile time. This is because, this level of control has some impact on the MCAT operation when user is accessing metadata. Many communities may not want this additional impact when they have no objection for everyone to see metadata. The option that turns this on and off is called the MDATAACCS. By default this option is turned OFF.
B.2. Softlinking of srbObjects to another collection This feature allows users to link their srbObject into a new Collection. In this mannat, a srbObject which is located in a Collection can be linked to one or more other collections. When this linking is done, no new physical copy of the srbObject is made (as done by the Scp copy operation). The new linked object can have its own user-defined metadata and annotations. But it shares the system-level metadata from the original copy. Hence access-control for this linked object is still with that of the original srbObject. None of the user-defined metadata are copied from the original to the linking copy. A user can copy the metadata if they wish using the copy-metadata functionality discussed below. To link an srbObject the user should have atleast a 'read' permission for the original object and at leasta 'write' permission in the collection they are placing the the link. The linked object will be visible through an Sls call in the new collection. Linking of a collection will be done in the next few patches. Linking of containers will not link the underlying files. One needs to link individual files in order to achieve this effect.
B.3. Streamlined Auditing Functionality The auditing functionality that is present in the Versions 1.1.8 and older are being phased out and a new type of auditing functionality is being introduced. This is to make the auditing to be of minimal overhead for many of the operations. The auditing level is controlled using environmental variable (AUDIT_LEVEL). By default AUDIT_LEVEL is 0 and no auditing is performed. When the AUDIT_LEVEL is set to 1 then operations on all datasets by any user gets audited and is written to special audit tables in the MCAT. Users can use SgetD to view audit information for datasets that they own. Auditing is done not only on operations of datasets but also for operations on collections, containers, resources and users. In later releases, we plan to introduce additional levels of auditing that may perform finer levels of auditing. Auditing is controlled using the runsrb script. So, a srbAdmin can turn on or off the auditing capability by stopping and restarting SRB after modifying the runsrb script. There is no necessity for recompiling the SRB for enabling/disabling auditing.
B.4. Locking and Hiding of srbObjects These two functionalities provide controlling of access to srbObjects on top of the access control lists and tickets. A owner can lock a srbObject using the SmodD command and following that user (including owners) access is disabled srbObject depending upon the types of lock that is applied. There are two types of lock that are provided. 'read' lock will read operations and 'all' lock will allow no operations (note that this is different in semantics from database-type locks). When locking a user has to associate a time-stamp with the lock. This timestamp provides the expiry date/time for the lock after which the lock is invalid. When a lock is expired any one with a write permission can release the lock. But when the lock is still alive only the lock owner (not the dataset owner) can release the lock. The lock only controls access to the srbObject and has no affect on the metadata for the srbObject. Hence reads, writes and modifies to the srbObject is still allowed. There are no collection-level lock in the current version. Lock is useful when one has created and modified a srbObject and is reasonably sure that the object will not be modified. Then locking will ensure that the file is not modified inadvertantly. Hiding a srbObject will make the object and its metadata 'disappear' from a users view. A hidden object seems as though it is deleted but it is still in the SRB and all its metadata are still stored. A hidden object can be unhidden and then it will become visible to users who are allowed to view it or its metadata. The hide and unhide operations can be performed using SmodD. Since a hidden object is 'invisible' to everyone, it is not seen by anyone including its owner when using Sls. A new option in SgetD allows owners to list all hidden objects in a collection. Another important point to remembr is that though the object is hidden it is still in the SRB and MCAT. Hence any attempt to create another object with the same name in the same collection will be disallowed. Hiding objects will be very useful when you want to take an object out of circulation but do not want to delete it from the collection. Hiding a container will not hide its objects. One need to hide individual object to achieve this effect.
SgetD with various options can be used to check various system metadata about data files and SmodD is useful for checking them. The following shows a few of the options and how they can be viewed. SgetD -f: parameter shown with SgetD how to set - dpin SmodD -P - dexpire_date_1 SmodD -x - dexpire_date_2 SmodD -X - dccompressed SmodD -z - dencrypted SmodD -y SgetD -e: - access_constraint SmodD -L - dchekcsum SmodD -k - dhide SmodD -H
B.5. Compound Resource and Compound Object Support MCAT has been augmented to deal with compund resources and compound objects. New APIs as well as data structures and tables have been included in this version for supporting the compound resource aspects of SRB.
B.6. Bulk Ingestion of Files Support This feature allows for bulk ingestion of files (directories of files) into a container. When this container is created, the directory structure gets translated into collections and sub-collections with files associated as srbObjects in these collection hierarchy. The building of this collection hierarchy is done using a bulk-load operation in the MCAT. For more information on the bulk load capability of SRB, please refer to the man pages for Sbload command.
B.7. Bulk Access Control API Function over all Objects in Collecions This functionality provides a means to provide permission to a user or a user group for all srbObjects in a collection using a single API. In previous versions of the SRB, this operation was done at per srbObject basis. Bulk access control will also work on collections in a recursive manner by performing the acces permission delegation to all sub-collections as well as to their descendents. Hence this is not a new functionality but an enhancement (speeding up) of existing function call.
B.8. Delete All from Container API This is another functionality that helps in speeding up already existing operation. In previous versions, deletion of srbObjects in a ccontainer was done at a per object basis which was very slow. With the current enhancement, one can delete all objects froma container using a single API call.
B.9. New Miscellaneaous Metadata attributes In order to increase the utility of the SRB system new attributes have been added to those already in the core schema. A new set of attributes holding the owner of a collection as well as its creation timestamps have been included. This allows not only for providing this information, but also helps in access control functions. At the srbObject level several new attributes have been added. The checksum is probably the most important one and allows the user to store and retrieve checksums of srbObjects. Hence, the integrity of any srbObject can be checked using the checksum. Two other new attributes store information about the encryption and compression (if any) of a srbObject. These attributes are basically character strings and can store any information regarding these operations. A suggestion is to store the methods of encryption (eg. RSA64bit) or comression (eg. Lempel-Ziv) that are performed on the object and this will provide a hint to the client application that it needs to post process the data after it has retrieved from the SRB or enable the SRB to apply appropriate proxy operations before sending the data to the client. Another new feature added is a set of attributes that enable pinning of replicas of a srbObject to their current physical location. Pinning of a copy of a srbObject to a particular place might be needed when one needs a file in a place for some operation in the near future (so that its access can be fast) and wants to prevent it from being purgd by some administrative daemon which does cache management using LRU type schemes. Pinning is not enforced by SRB by itself but can be used by such daemon which are doing cache management on a local file system. Obviously there is some chnace that this can be abused by the users who might pin every one of their files; but in such a case, one can view these as non-mandatory pinning and a higher logic will be applied to remove even pined files. Hopefully, in a cooperative situation, pinning can ideally be used for improving data access performance by using judicious data placements. Pinning has an expiry timestamp associated with it, after which the pin becomes invalid. The owner of the pin can at an time unpin the srbObject. Two other attributes that are useful for storage recalamation are also added in the new version: expiration_date, expiration_date_2. The second attribute can be viewed as providing a timestamp after which the srbObject can be put into archival storage and all its cache copies can be removed. The first expiration timestamp can be used as a time when the srbObject can be removed completely from the SRB system. These are very useful when one has intermediate results that one would like to have for some time but want it to be automatically purged after some time. These are also useful for digital library management where one would like to purge from active disks collections of files after some time period. Two other atributes that are added are useful for SRB itself. One is the Segment number which store the segment number of a srbObject that is split into more than one piece and stored as individual files. The SRB can chain these segments and provide a contiguous piece back to the users. This is helpful when the file size is too large to be held in a particular file system. Refer to the section on MSS for additional details how the segmentation is used. The other attribute is called the version number and can store the version number of a srbObject. AT the current time this is not utilized by any of the API or Scommands. In the near future we plan to provide a simple versioning capability in the SRB and will utilize this attribute.
C. Improvements to User-defined Metadata Functionality SRB Version 2.0.0 has a few extensions to the user-definable metadata capability of the SRB. These range from entirely new set of user-defined attrbute sets to additional querying functionality to metadata ingestion assistance. C.1. Metadata Extraction Functionality This facility allows a user to define a method for extracting metadata from a srbObject and assobciating them to a (possibly different) srbObject at the SRB server level. For example, if one has stored a DICOM image in SRB, then one can extract the metadata from this image file and associate the elements as user-defined metadata for the DICOM image. This will allow susers to query on the fields of the DICOM attribute schema and discover images of their interest. In order to do this, a user associates a style-sheet type program written in the T-language and registers this as a srbObject with user-defined metadata for this T-language template stating that it is a metadata extraction method for DICOM image data type. One can associate more than one method to a data type as well as associate the same template for more than one data type. Once this is done, for any DICOM image file, the one can invoke the metadata extraction function and then the metadata will be extracted from a source file and associated with the target DICOM file. Note that the metadata extraction can be done on one srbObject and the resulting metadata assciated with another file. This is needed because in many cases, one may store metadata in an XML file.
C.2. Copy Functionality for Metadata and Annotations This feature allows a user to copy metadata from one srbObject or collection and associate them with another srbObject or collection. Any old metadata that is associated with the target srbObject or collection is not deleted. The user prforming the copy should have at least 'read' permission for the source and should have at least 'write' permission for the target. Note that copying from a srbObject to a collection and vice versa is allowed. This functionality is also provided for annotations where one can copy annotation from a srbObject to another srbObject. Copying of annotations to and from collections will be provided in later patches.
C.3. Additional Query Capability for user-defined metadata In version 1.1.8, one can query user-defined metadata only using a single attribute at a time. This, though useful, was extremely time consuming and can lead to lot of post processing. In order to alleviate this difficulty, this version has been enhanced to query upto five metadata attributes at a ime. Our user requirements seem that five is enough for the present. If there is a necessity for any project to query more than five attributes, the SRB developers will accomodate that request. The five attributes are queried conjunctively (i.e., using the AND operation). In the near future, we plan to release where one can have a choice of conjunction or inclusive disjunction in their queries.
C.4. Annotations for Collections As per user request, we have added functionality where one can annotate collections also. The operation is similar to that provided in version 1.1.8 for srbObjects.
C.5. User-defined Metadata for Resources and Users A new set of attributes have been defined for storing user-defined metadata for resources and users. This will allow one to store arbitarary metadata for resources and users. These will be very useful for administrative purposes as well as for SRB data access purposes. For a resource, one can store information about its current load, delay, down time, preventive maintenance schedules, and other operational characteristics. For users, one can store their profile information and information regarding their usage model and such. At the current time, these metadata attributes are not used by the SRB system. In the near future, we plan to use them to provide additional functionalities.
D. Administrative Support Improvements have been made in MCAT adminsitration. One of the main change is that the administartion of MCAT can be done from the client side (as opposed to server-side admin tools available with version 1.1.8). D.1. Client-side Java GUI for MCAT Administration The Java tool for MCAT/SRB administration has been enhanced and has been made to operate from the client-side. This will allow one to administer SRB for registering users, resources, etc from any Java-capable system. In version 1.1.8, administration was tied to the system that ran the MCAT. More information about the new administration Java GUI can be found in README.MCAT.ADMIN.
D.2. Client exposure to Administrative APIs for Admin GUI The development of the client-side GUI for SRB/MCAT administration required that these APIs, which were previously available only on MCAT-enabled servers, be available at the client-side. In version 2.0.0, all adminsitration APIs are available at the client-side. See web/SRB.htm for more information.
D.3. New Admin Functionality for remove: users, resources, locations,domains. SRB version 1.1.8 had capabilities for registering entities such as users and resources but lacked any means for deleting them. In version 2.0.0, we have alleviated this difficulty. The new Java Admin GUI can be used to delete users and resources. The deletion will go through only if no associated references are there for the entity being deleted. For example, if one is deleting a user, then no srbObjects or collections should be associated with that user. Such checks are automatically performed and a deletion is allowed only if all conditions are satisfactory.