This project is read-only.

Just me? Indexed slower than non-indexed

Feb 5, 2010 at 5:35 PM

I ran the i4o Demo this morning on the latest build with the following parameters:

Student Count: 3000000
Student Name: 999

I was so surprised that I ran the test several more times with no changes:

Returned Students (with index) = 3833
Query occured (sic) in 3408ms!
Returned Students (with no index) = 3833
Query occured (sic) in 225ms!

Note that "occured" should be spelled "occurred".  I went ahead and flipped the order of the tests (always a good thing to do) and got the following results:

Returned Students (with no index) = 1521
Query occured (sic) in 189ms!
Returned Students (with index) = 1521
Query occured (sic) in 32ms!

What?!  Can someone replicate on their end?  On the other hand, the join test does show dramatically improved performance.

Cheers,
Tim

Feb 5, 2010 at 5:58 PM

Setup the test so that it runs index/no-index/index/no-index.

Student Count: 1000000
Student Name: 888

Run 1:
With index - 32ms
With no index - 66ms
With index - 7ms
With no index - 66ms

Run 2 (restart, flipped):
With no index - 64ms
With index - 32ms
With no index - 64ms
With index - 7mx

Student Count: 3000000
Student Name: 888

Run 1:
With no index - 183ms
With index - 31ms
With no index - 187ms
With index - 8 ms

Run 2 (restart, flipped):
With index - 34ms
With no index - 180ms
With index - 9ms
With no index - 180ms

It was probably just a fluke - maybe my AV was running or something.  Please disregard my previous message (except for the part about "occured").