Archives of the TeradataForum
Message Posted: Fri, 26 Mar 2004 @ 19:03:31 GMT
| Subj: | | Re: Aggregate Join Index & Fallback |
| |
| From: | | Victor Sokovin |
Carrie,
| | I haven't tested this condition either, but I checked in the Database Design manual V2R5.1 (page 12-6) and the text does suggest that a
join index with fallback WOULD be used if an AMP is down. | |
| | "If you do not define fallback for a join index and an AMP is down, then the following additional criteria become
critical: | |
| | 1. The join index cannot be used by the Optimizer to solve any queries that it covers | |
| | 2. The base tables on which the join index is defined cannot be updated." | |
| | If the join index is being used for partial covering (which results in joining back to the base table rows) the optimizer would not be able
to use the JI if the base table was not also fallback protected, but it sounds like a JI that covers the query completely would be used. | |
Let's see ... This text comes after the following phrase:
"You can define fallback for join indexes. The criteria for deciding whether to define a join index with fallback are similar to
those used for deciding whether to define fallback on base tables." [on the same page]
Together I read this as if they are discussing the additional criteria one has to have in mind when deciding NOT to define fallback on JI.
In total there are four combinations:
base table fallback Yes/No;
JI fallback Yes/No.
Donald's example is of the "No, Yes" type but there could be another situation: "Yes, No". The latter is especially covered by the second
point (as above). Although the base table is fallback protected, it won't be updatable if an AMP fails because the JI is _not_ fallback protected.
So, the base table won't have its full functionality and this becomes an additional point of concern when selecting data protection policy in the
presence of JI. That's how I read it.
Clay, could you perhaps help with some test results?
Thanks,
Victor
| 
