performance issue--hash file?

[dnat]

Hi,

We recently made some code changes to an existing job.

The hash file had one key and we added one more key(so now it has combination key) now. After this code change we are having tremendous performance issue. it is processing at 3 rows/sec where we are dealing with 150 millions of records. Earlier it used to process at 600 rows/sec.

Can anyone tell me, whether this could be a problem(adding one more key to hash file)

[ArndW]

Hashed files only have one key, when you define multiple keys the actaul key used is a concatenation of the fields, separated by a special character. You have not stated whether your problem is read speed or write speed. If the write speed has slowed down, then it is due to the number of bytes in the composite key. If your read speed has decreased, then it is probably because you are reading just one part of the composite key and the hashed file read now has to do the equivalent of a full table scan for the partial key. There are options available - one of which is declaring an index on the partial key column but that will slow down the write. Do you load this hashed file to memory?

[dnat]

Hi,

We are actually reading the hash file as reference and updating in the same file. after adding the key the performance has degraded extremely. Is there any way to overcome this situation

[ray.wurlod]

Use a single column key, perhaps made up from the concatenation of the two fields plus a separator character.