Posted: Mon Nov 27, 2006 1:27 pm
There is primitive help in the Hashed File Calculator - it all pops up in message boxes. However, the help is about how to use the tool, rather than about the underlying calculations.
Otherwise you need to consult the UniVerse manuals.
The important measure is the number of bytes per record. In your case this is 347 (how did you calculate this?) Your calculation was based on the length only of the key fields, which is not how it should be done. The first approximation for MINIMUM.MODULUS, or the modulo setting for a static hashed file, should be
(total records * average record length) / (group size * 80%)
= (402539 * 347) / (2048 * 0.80)
= 82525
If you did not include physical storage overheads, add 14 bytes per record, so you would use 361 rather than 347 bytes/record in the above calculation, yielding a figure of 88695 as the result.
You do not need to increase GROUP.SIZE, since your average record size is small to medium, and ANALYZE.FILE reports that you have no oversized records. You do not change hashing algorithm, because your keys do not form an integer sequence. SPLIT.LOAD, MERGE.LOAD and LARGE.RECORD should also be left as they are.
Otherwise you need to consult the UniVerse manuals.
The important measure is the number of bytes per record. In your case this is 347 (how did you calculate this?) Your calculation was based on the length only of the key fields, which is not how it should be done. The first approximation for MINIMUM.MODULUS, or the modulo setting for a static hashed file, should be
(total records * average record length) / (group size * 80%)
= (402539 * 347) / (2048 * 0.80)
= 82525
If you did not include physical storage overheads, add 14 bytes per record, so you would use 361 rather than 347 bytes/record in the above calculation, yielding a figure of 88695 as the result.
You do not need to increase GROUP.SIZE, since your average record size is small to medium, and ANALYZE.FILE reports that you have no oversized records. You do not change hashing algorithm, because your keys do not form an integer sequence. SPLIT.LOAD, MERGE.LOAD and LARGE.RECORD should also be left as they are.