Performing total sort

[zulfi123786]

I was trying to implement a total sort using Range partitioning and hence as a prerequisite created a range map file using write range map stage. Out of curiosity wanted to see what this file contained but it turned out to be a binary file, any documentation which states what layout/format this file follows ?

Also during the implemention of range partitioning does datastage ensure that key values are assigned in ascending order to nodes as they appear in the configuration file ? if so then ordered collection could be used else have to go with sort merge collection which per my understanding would sort data across partitions (similiar to sorting on single node)

Thanks

[ray.wurlod]

Structure of the range map is not documented in the public domain. You can be fairly certain that it contains the limit values for each node.

Range partitioning does not perform a sort, though it will preserve any sorted order of the partitioning key that happens to exist on the input link. This is purely an artefact of the FIFO nature of record processing. Therefore, if this is your scenario, then an Ordered collection algorithm will be apposite.

[zulfi123786]

what particularly interest me is how the key boundaries are defined per each node, which node gets the keys with least values when the range map is created and so on as this would decide the link ordering during ordered collection when the actual input is not sorted to achieve a total sort.

Also wondering how datastage would dynamically adjust if the range map is created in x nodes and the same is used in range partitioning in a job running on y nodes. Would have tested this myself by dont have the access to datastage right now

[ray.wurlod]

Your proposed scenario of different number of nodes for range map will fail.