Which is faster for incremental loads: Partial Reload or Full Reload
I have a QVW that serves as the "data model" for a downstream visualization app (binary loaded) in which I am pulling incremental QVX files from a datasource (e.g. from Salesforce.com). Imagine 5-10 tables in memory with 2-20 columns each and 10 - 1,000,0000 records each.
I am wondering which method will be faster and provide greater scalability as the number of rows and the number of tables in the data model increase: traditional "Reload" with the WHERE NOT EXISTS(keyField) OR a partial reload - also with WHERE NOT EXISTS(keyField).
Is this dependent on data complexity (e.g. number of columns, cardinality, numbers of incremental rows, etc.)? Or, is it simply always better to use the Reload with WHERE NOT EXISTS()?
Any experience - and especially data-backed benchmarking! - would be appreciated.
The partial reload option would require code patterned after the following: