DSXchange

Hi,
May I know why not to use Server shared container on MPP environment. Won't Datastage have a common repository accross all the CPUs?

I would assume because they just don't work. From the docs:

You can only use Server Shared Containers in this way on SMP systems (not MPP or cluster systems).

Others may know the gory details of the 'why'. [shrug] Or ask IBM.

I think, the reason lies in the very architecture of the two systems and the nature of a shared container. SMP uses a single shared memory irrespective of how many cpus are connected to it. The result of the shared container resides in the shared memory and whatever cpu requires the shared contrainer, requests for it.
Where as MPP's are, briefly, many computers hooked up for a massive single unit. It might not be able to work in the same way as the shared container was originally designed to work in and hence the limitation.

I might be wrong, but need someone to confirm it.

But in one of a previous post, Michael has posted that its working in MPP as well.
Just curious to know, will datastage store its repository in each CPU or a common repository will be maintained.

Can you post the link to Micheal's post?
The manual says pretty clearly that it wont work, or atleast not supported. I wouldnt go for something that might break some day.

Here it is. Though he haven't given any strong asertion about that, it seems it working.
viewtopic.php?t=85355&highlight=mpp+shared+container

Even you can find Ray is pointing to his post in this.
viewtopic.php?t=85353&highlight=mpp+shared+container

Go for it then. As I said before, I wouldnt go for something thats a trial, or might break someday. If it does, you cannot even go back to IBM, as they will point out to the same line that Craig quoted, that it is not supported.

I am not going to run it anyways, I just want to know why its recomended in such way. And like to know the architecture of Datastage on MPP.