Investigate if its meaningfully more performant to do a direct read of the samples collection

XMLWordPrintableJSON

    • Type: Task
    • Resolution: Unresolved
    • Priority: Minor - P4
    • None
    • Affects Version/s: None
    • Component/s: None
    • None
    • Query Optimization
    • None
    • None
    • None
    • None
    • None
    • None
    • None

      Instead of setting up a collection scan on the write path as we did in SERVER-130446, it may be faster to do something like

      auto cursor = collection.getCollectionPtr()->getCursor(opCtx);
// Seek to first page
auto record = cursor->seekExact(recordIdForId(minId));
std::vector<BSONObj> pages;
while (record) {
    // Check if we've exceeded max bound
    if (record->id > recordIdForId(maxId)) {
        break;
    }
    
    pages.push_back(record->data.toBson().getOwned());
    record = cursor->next();
}

      We should create a google benchmark to test this

            Assignee:
            Unassigned
            Reporter:
            Militsa Sotirova
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

              Created:
              Updated: