Skip to Content
author's profile photo Former Member
Former Member

Manually managing Delta Merge

I look forward to someone specific answering this one. :)

Due to the "double buffer" concept, a table is locked during the delta merge process for only a very brief time. Table writes during delta merge go to "Delta 2", which - as a delta store - is as optimized for writes in the same fashion as the 'original' Delta 1. (Reference:

I'm finding a lot of "experts" note that manually disabling delta merges before initial loads, completing the load, and re-enabling after the load - will increase "performance".

Perhaps I'm missing something obvious, but when people speak of performance - they're usually referring to runtime - and it's not clear why manually altering the delta merge process has any impact on runtime.

Memory - sure, I understand that the delta loads main to delta to complete, and thus there may be valid reasons for managing memory (and thus delta) manually - but not clear on why this should impact runtime performance.

To be clear, these "experts" may well know something I don't, and even I like to refer to myself as an "expert".

Or perhaps more likely - I just like to put certain terms in double quotes. ;)

Add comment
10|10000 characters needed characters exceeded

1 Answer

  • Posted on Dec 20, 2016 at 07:04 AM

    And this would be my approach to an answer:

    Merging right. Some tests with data loading in HANA from Down Under.

    Short version is: yes, it is beneficial to manually handle the data load, because you can use your knowledge about when your loading is finished to make a better decision when a merge should happen.

    Long version is a lot more complex and required that I drew pictures... from that you can tell that it got nasty :)



    Add comment
    10|10000 characters needed characters exceeded