Hi Tom,

We have data coming into our application from different entities/agencies and being loaded into several different tables. Each transmission contains data that belongs to a unique entity. We can also receive a new transmission for each entity, in which case we need to delete the previous transmission's data. My thought was to partition each table by that entity, i.e. use list partitioning, since we also get an entity id in our transmission. All our subsequent use of this data is always divided at the entity level, so the partitioning works out nicely. In the event we receive a new set of data, I can simply truncate the existing partition and reload the new data. However the size of the incoming data can vary wildly, e.g. we can have data for one entity that is in the low thousands, and for other entities that can go into the millions. My concern is that I will end up with partitions that will not be sized equally, so some partitions will be a few megs, and other partitions can be several gigs. Note that we do not have any indexes on the tables being loaded.

Is this something I need worry about? Any ideas/best practices for such a requirement? I also toyed with the idea of using a bitmap index on the entity id, so that each query against these tables will be able to use this index. At any given time I do not anticipate us having more than 20 entities. However with the variation in the size of each entity's data, I do not think I will gain much from a bitmap index, especially for the entities with larger data.

Thanks as always.

Kashif