Grid Wrapper for WebSphere Information Integrator
A tool that allows Websphere Information Integrator to access data grids using OGSA-DAI data grid middleware.
Date Posted: August 16, 2005
|
|
 |
|
|
What is Grid Wrapper for WebSphere Information Integrator?
Usually, applications must be grid-enabled by, for instance, using OGSA-DAI (Open Grid Services Architecture -- Data Access and Integration) APIs, and that can impact the applications too much. Grid Wrapper for WebSphere Information Integrator enables applications that currently access DB2 (or any other RDBMS) to transparently access a fully-distributed database over the grid with minimum changes.
A registry (either XML or relational) key to this wrapper enables the dynamic aspects of the grid to be fully used. For instance, the registry allows the system to cope with failure by finding the next available database on the grid with similar information; finding this database enables the application to fetch results in parallel from multiple databases or replicas on the grid. Other access patterns, such as fetching partial results and using load balancing to fetch from different databases, are also possible.
How does it work? This technology uses the newly-released Java wrapper SDK, which is part of Websphere Information Integrator 8.2. By getting metadata information from a registry (XML, relational, or OGSA-DAI), the wrapper will make decisions to fetch the data from the proper database or databases. This is totally transparent to the DB2 user. For instance, before the wrapper, the query looks like this: "select * from mytable." After the wrapper is installed, nicknames created on mytable, and the registry loaded with information such as "mytable is in database 1 but also replicated in database 2 and 3", the wrapper can make the decision at run time, just before shipping the query to the remote database in order to actually ship it to Database 1 by default. If Database 1 is down (an error code comes back) the wrapper can automatically try Databases 2 and 3. The wrapper could also decide to fetch data from both databases if the table is partitioned on these two databases. Further possibilities are described in the included tutorial.
Other access patterns are implemented in the wrapper; for example, third-party delivery (using OGSA-DAI) allows a request to be submitted to multiple databases on the grid and enables the results be shipped to a gridFTP server instead of directly back to the application. This saves bandwith, especially for grids that deal with large amounts of data.
|
|
|
| | About the technology author(s):
Patrick Dantressangle worked for eight years within IBM's information management group on projects including Net.Data, the 1998 Nagano Winter Olympics Web site, the initial implementation of the DB2 stored procedure language in DB2 UDB 7, and the DB2 UDB migration toolkit (MTK). Lately, Mr. Dantressangle has been working on grid computing projects, including OGSA-DAI, SIMDAT, and others.
James Magowan worked for eight years on IBM's high performance computing, large SP installations, and Linux clusters. Recently, Mr. Magowan has worked on grid-computing projects, including European DataGrid and OGSA-DAI; he has served as chairman of two working groups in the Global Grid Forum; and he has been working with Web services and data grids.
Adrian Lee, of the Emerging Technologies Services team at the IBM Hursley Development Laboratory, works on integrating existing IBM data management products with emerging technologies such as grid computing. Mr. Lee has 12 years experience of providing technical consulting on IBM's data management products. His total career in the IT industry spans 21 years.
Fabian Bannwart worked as a software developer in the finance industry for several years and participated in academic projects concerning privacy enforcement and the axiomatic verification of bytecode.
|
|
|
| |
