DSXchange

Is it possible to split a Dataset through Unix and use the split files as inputs to a DataStage PX Job ?

if so, what are the precautions that one needs to take ?

Umm... What do you mean by "dataset" here? In the PX environment we usually refer to a dataset as something created by DataStage PX. Also, a fileset is something similar, the main difference being that a dataset is in memory to the greatest degree possible, while a fileset exists in persistent storage.
Can you please clarify? If it's a PX dataset, there's no need to split it; the PX control file (the one with a ".ds" suffix) contains the information about how and where it's split.
If you mean a data set as something like the result set of the query, then simply running the query within the PX environment will cause the data to be "fanned out" over the available resources.
All this is well explained in the Parallel Job Developer's Guide.

If i understand Inquisitive correctly, you want to split a DataSet just like a flat file on Unix.

No, you cannot do this. Just because the data in a Dataset is already split and stored in nodes.

If you want to distribute your data into various datasets, do it in the job itself, by having multiple output links from your last stage and writing into more than one DataSet stage.

Hope i answered your query.

Regards,
Sandyla.

You can split dataset in two ways

1) create a small job which reads the dataset and writes to another dataset. When you run, give the correct $APT_CONFIG_FILE which tells how to split the file (you have to create correct config file)

2) If you are familiar with unix, then use orchadmin command.

Siva
(for Luke)

Inquisitive wrote:Is it possible to split a Dataset through Unix and use the split files as inputs to a DataStage PX Job ?

if so, what are the precautions that one needs to take ?