When installing additional remote engine tiers to function as compute nodes, can they be installed onto a different platform than the primary engine tier machine / conductor node?
DataStage requires all machines serving as conductor and compute nodes (engine tier machines) must run on the same platform and ideally should be at the same OS level to ensure code compiled on one machine can run without errors on all machines.
The primary reason for this restriction is that the parallel / C transformer stages are compiled C code that we compile on the DataStage server when the user requests a job compile. The C compile occurs on the primary engine tier machine which is also the conductor node. That compiled C program is then either copied or shared (via shared mount point) to all the other remote engine tier machines.
Transformer stages compiled on one operating system such as AIX are not executable on other operating systems such as Linux. Thus every job with a transformer stage would fail unless it runs only on the conductor node.
An additional reason occurs due to the data passed between conductor and compute nodes is not sent in a platform independent format. Since some platforms format data differently (byte order, etc) than others, this would also cause a problem.
Running engine tier machines for a single DataStage installation on multiple platforms is not currently supported.