Anbody used Hash partioning? - what about queries not using partition key

Question

Hi there,
We have a table, will have over 3 billion rows when loaded and wish to implement partioning.
Brief structure
cal_id -- relates to date of transaction
prod_id
cust_id,
sales_qty
The id's are all surrogate id's/dimension keys
About 80% of queries expected to use by on_sale_date of product rather than date of transaction hence considering using hash partition on product_id.
Queries will hit > 1 partition in general and also some e.g 15% of queries wil be by date of transaction.
Qusetions I have - how will these 15% of queries be affected. I expect will be poorer performance-wise because of partioning but becuase of size of table and 80% queries by product have to use partition.
Also will partition-pruning work if queries use between rather than = operator
Anbody used hash partioning - keen to hear other people's ideas/experience with hash partioning.
Many Thanks

Answer 1

Hi,
When partition pruning is not used, then that depend on the access path.
If it does a full table scan, then there is nearly no overhead to full scan all partitions (compared with full scanning the same non-partitioned table)
If it is an index access and if indexes are local, then all indexes must be read. But with global index, again, there is no overhead in comparison with a non-partitioned table (except that indexes are a bit larger when global on partitioned table)
However can be expected to partition even when reading all partitions, especially with parallel query.
the between is not optimal with hash partitions, s it is hash partition, not range partition.
Regards,
Franck.