Archives of the TeradataForum
Message Posted: Mon, 17 May 2004 @ 14:07:13 GMT
| Subj: || || Re: Are Numeric Keys and Indices more efficient? |
| From: || || Geoffrey Rommel |
| ||I was always told that these were more efficient (i.e. faster) than character fields. Is this still (or was it ever) true? If it does
make a difference, has anyone seen any reliable statistics for how big of a difference it makes?|| |
This question came up recently here. I did not conduct a systematic test, but I did test one scenario.
To compare the speed of decimal(13) vs. char(15), I copied the A and B tables to a work area twice. One copy of each table had a decimal(13)
primary index; the other copy had char(15). Otherwise, they were identical. All tables had the same primary index. I then joined them. The
results were as follows:
Elapsed time Total CPU Total I/O's
decimal(13) 5:28 1,931.18 1,950,115
char(15) 5:21 4,043.43 8,818,803
These results came as a surprise to me. I guess numerics are a lot better after all!