SEARCH
0-9 A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
Prev | Current Page 212 | Next

Robert Wrembel and Christian Koncilia

"Data Warehouses and Olap: Concepts, Architectures and Solutions"

These components look like a
processUnit, but on a higher level they have a termination status; their dependency
rules are structured in simple trees and not in graphs. Each component can then
have an associated recoveryProcessUnit, so called because it keeps the undo code
(e.g., truncate partition).
The synchronization engine keeps track (on the DBMS) of the termination status of
a single processUnit, so the engine can evaluate the status of each component and
of the entire ETL process itself. At the end, we have a status table that says if some
failure has occurred (the process needs recovery) and which component has failed.
With this information, we can recover only the failed components and, which is
08 Adzic, Fiore, & Sisto
Copyright ?© 2007, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of
Idea Group Inc. is prohibited.
important, in an automatic way. Simplifying, the synchronization engine, running
in recovery mode, has to do a reversal of the component status and then start only
the failed components (or to be more precise, the processUnits of the failed components).
In practice, there are some further complications; a component in recovery
mode can depend on a previous one that has not failed but must be rerun because it
does an initialization job (and so we have introduced the onRecovery dependence);
then the process running in recovery mode must always use a point-in-time date
and so on with other little complications.


Pages:
200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224