Hash file as a global reference

[paddu]

Hi there,

I have a requirement where I need to use a hash file as global reference in several jobs out those jobs I have to parallel atleast two or more.

So two or more jobs using the the same hash file as a reference(lookup) as well as I need to insert new records into that hash file in order to keep it current. How good this approach will work and is performence an issue?

Will hash file work based on some file lock protocol for reference vs update with different processes running at the same time?

If this approach is not the right one, what would be the best way to go when I have to use the same dataset as a global reference in several jobs where I also need to update the hashfile.

Thanks in advance.
Paddu

[kcbland]

Run the jobs that update the hash file first, then all of the other jobs can access a static hash file. Using locking mechanisms will cause a lot of headaches as jobs will stall waiting on locks, plus you probably will end up hanging your jobs when you aren't releasing locks quickly enough. This is a bad approach, take the recommendation stated.

[susanm]

Hello Paddu,

You can also investigate the use of hashed file cache sharing.

There is a check box on the properties page of the job where you specify that you want to share the hashed file cache. This will allow several jobs to share the hashed file cache for both reading and writing.

Before you can use this option you need to update your uvconfig file to set some caching tunables.

For further reading refer to the DataStage Disk Cache Sharing Guide.

Susan