Hash File Options

narasimha · Post by **narasimha** » Thu Apr 07, 2005 3:04 pm

Hi,

I am trying to run a job which creates a HASH file from a table, there arearounf 12 million recordsin the table. In the log the status shows as waiting for the job XXX to start. The status shows as running only after a period of 30 minutes. I am using the default hash file options i.e. Type 30(Dynamic).
1.What are the best options followed for faster completion of the job?2.Should the Allow Stage write cache be enabled
3. Is it better to delete file before create
Any help appreciated

chucksmith · Post by **chucksmith** » Thu Apr 07, 2005 3:34 pm

If you are just writing to the hashed file, then enable write cache.

narasimha · Post by **narasimha** » Thu Apr 07, 2005 3:41 pm

Thanks for the info
I want to improve/reduce the time taken for writing into the hash file. What changes can i make to the Hash files option???

ray.wurlod · Post by **ray.wurlod** » Thu Apr 07, 2005 5:00 pm

The write cache is good. What you get is writing to random locations in memory rather than writing to random locations on disk (within the file structure, of course). Once all rows have been written, the cache is flushed to disk in group order, which is maximally efficient I/O.

You can preallocate disk space by specifying MINIMUM.MODULUS when creating the hashed file. You can also preallocate disk space by specifying a static hashed file (type 18 for example). The latter course is indicated where the number of rows is known in advance and unlikely to change - or unlikely to change much - over time. Preallocating disk space (that is, preallocating the number of groups) will be beneficial even if write cache is used.

You can not use write cache if you're also reading from the hashed file while it is being written to.

narasimha · Post by **narasimha** » Thu Apr 07, 2005 5:12 pm

Thanks Ray

The number of rows cannot be predetermined (So i wont be able to use Static Hash File). For preallocate disk space by specifying MINIMUM.MODULUS do I use the standard formula's or is there any other way we can optimize the performance?

gpaitch · Post by **gpaitch** » Thu Apr 07, 2005 11:43 pm

In the Utilities direcoty of the ds install disk there are some unsupported utilities there. One of these utilities is hfc.exe. This is the Hashed File Calculator, which will provide the modulus based on the number of bytes for each row.

subramanya · Post by **subramanya** » Fri Apr 08, 2005 12:27 am

Hi,
If we enable "Delete file before create" option in the hash file and run the job repeatedly, we are getting the following error "Unable to delete. The file is allready in the VOC list". This happens with some of the hash files. Can anybody explain Why?

Thnx
Subramanya

chulett · Post by **chulett** » Fri Apr 08, 2005 12:43 am

Define "repeatedly"? I'm guessing the delete is failing because it's actually still in use when the attempt is made.

Are you sure it doesn't say "Unable to create" instead?

ps. You really should have started a new post on this specific topic, rather than piggy-backing on the end of an existing (but at least semi-related) thread.