DSXchange

Suppose a project has server jobs only, is there any benefit implementing the same in Information Server Blade?

Is there any manuals or documentation available for the same? I could not find any in the IBM website except for a single page.

If possible, please share your experiences if you had implemented Information Server Blade.

Thanks for your time.

What is Blade?

The Information Server Blade comes with three blades (three separate motherboards) with dual core CPUs on each. The first blade comes with Metadata Server and one dual core processor, the second with DataStage head node and the third with DataStage child node (sounds like a Gilette ad) both with twin dual core processors.

The advantage for Server Edition customers is that the metadata server and DataStage comes already installed and running, the problem is the difficulty in getting DataStage Server jobs to run against and fully utilise the twin DataStage blades and the twin CPUs. It comes with Tivoli Workload Scheduler for running jobs across the blades but I don't know if that is enough to run multiple DataStage Server jobs on available blades. I assume you would have the standard problems with Server jobs on multiple CPUs - bottlenecks and underutilised resources. Anyone had any experiences with DataStage Server Edition across multiple blades?

Thanks Vincent for the information.

I read that Tivoli workload scheduler does load balancing and failover mechanism between the blades. Does this means that EE configuration file is generated dynamically for each job invocation by the Tivoli Workload scheduler?

That's the idea. With the Tivoli software you don't have to keep tweaking the config files, they are generated on the run. I don't know how pratical this is if you dont know how many rows a job is going to process and you don't want small jobs running too many instances. Sometimes you want to hard wire a job to always use a single node or blade config file.

Vincent - a question...

Are the config files actually generated on the fly or just selected (as appropriate) from manually created configs that have been defined to fit the situation?

They use the same dynamic configuration files that any grid configuration would use. The basis is a "standard" configuration file, and there are some environment variables that need to be set. I can not remember off the top of my head which ones they are.

and there are some environment variables that need to be set. I can not remember off the top of my head which ones they are.

APT_GRID_COMPUTENODES?

Interesting discussion - I have no knowledge of Blade but to answer the original question, it sounds like the advantages of Blade are the preconfig of Metadata Server and Tivoli. The usage of multiple CPUs by Server should be the same as any other multi- CPU box. You can run multiple DataStage jobs at the same time and use many CPUS or you can multiinstance jobs and run many instances. There are ways to 'psuedo dynamically' control the number of instances run by using sophisticated job control (i.e. Ken Bland's framework) and parameterizing the number of instances.