Skip to content

Performance issues in viewing rivers on specific systems #8

@scripting

Description

@scripting

We've been working in various email threads trying to figure out why performance in viewing rivers is so slow on feedland.com.

Here's a copy of the last email I'm responding to in the next comment.

I am fairly confident that a new index with columns flDeleted, feedUrl, pubDate (in that order) would improve this query as-is. 5 minutes sorting 1.08M rows seems incredibly long, even if the query is sub-optimal. It's difficult to speculate what might be causing poor performance.

I'd like to experiment using a JOIN instead of a sub-select. I believe a join would be able to apply the pubDate condition to the entire recordset. Whereas the in() is: first retrieving all the records for all the feedUrls returned by the subselect, putting those in a temp table, and sorting that temp table.

A snapshot of the database would help validate this.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions