Home/ Questions/Q 7938297
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-06-03T22:46:13+00:00

I executed an update statement along the following lines yesterday: UPDATE MainTable Set SomeField

  • 0

I executed an update statement along the following lines yesterday:

UPDATE MainTable
Set SomeField = SubsetTable.SomeField
where MainTable.MainTableKey = SubsetTable.MainTableKey

where SubsetTable is a subset of MainTable and has the same Primary Key field. MainTable has roughly 200m records, SubsetTable has 5m records. MainTableKey is a GUID.

Both of these table have a clustered index on MainTableKey.

When I executed this query the first time it took a whopping 14 hours.

Then I added a non-clustered index to MainTableKey on both tables. Now it takes 30 minutes.

Does anyone have any ideas on why the performance gain would be so dramatic?

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    I bet if you look at the execution plans:

    The first query is a merge join which involved reading both tables completely.

    200M rows + 5M rows = 205M rows.
205M rows / 14 hours = 4067 rows per second.

    The second query is a nested loop join which reads the whole small table and seeks into the large table for each small table row.

    5M + 5M rows = 10M rows.
10M rows / 40 minutes = 4166rows per second.

    That these rates are approximately equal, supports my theory about which rows are read.

    You don’t have to guess: Run the queries with SET STATISTICS IO ON, and/or look at the execution plans.

    • 0