20 Sep
2019
20 Sep
'19
11:42 a.m.
Hi Sergey!
In the best case scenario, I think this could enable InnoDB to use its page directory to skip past some of the keys, improving the performance from O(n) to O(log n).
Agree. If the scans being merged produce data with non-overlapping rowid ranges, then things could be sped up. I'm just wondering how often this is the case in practice. Do you have any thoughts this?
In case of a ecommerce solution using a Faceted Search, it happens really often! Usually users are moving away the Search part from MySQL to ElasticSearch / Algolia because the index merge performance is really poor. If we have this missing feature, it could improve a lot performances of such cases! BR, Jocelyn Fournier