Team,
I do not know anything about parallel extender. In our comapany till now they are using 7.1 without parallel ext. And now I am asked to gather information on it. They want me to create a prototype. Can you please let me know how to go about it.
Thanks
Parallel Extender
Moderators: chulett, rschirm, roy
</a>You also need the software!
I noticed that the earlier posts all assumed you actually have the software. Is your company was running on Enterprise Edition or Server Edition? Enterprise Edition does contain both the Server and PX versions of DataStage. It is considerably more expensive than standalone Server edition.
If you are running on Server Edition and need to try out PX you can contact your local IBM rep to see if they can provide a 30-day evaluation copy of Enterprise Edition. Ascential used to provide these, don't know how easy it is to get them from IBM now.
If you are running on Server Edition and need to try out PX you can contact your local IBM rep to see if they can provide a 30-day evaluation copy of Enterprise Edition. Ascential used to provide these, don't know how easy it is to get them from IBM now.
ewartpm - a DataStage PX project includes all of the server functionality; so the answer to your question is Yes, you can import your existing server jobs into PX.
But in order to use the full power of PX you will need to re-write those existing server jobs, and the development time for converting an existing job is from Server to PX is not that much less than it took to write the job initially.
But in order to use the full power of PX you will need to re-write those existing server jobs, and the development time for converting an existing job is from Server to PX is not that much less than it took to write the job initially.
Rewriting the Server programs can be a daunting task, especially if you have a lot of them. However, rewriting a program that is known to work into a new format (ex. from Server to PX) is probably one of the best ways to learn the tool.
We never used Server, but we had Torrent in house here. All of our Torrent code had to be rewritten as PX jobs. The DataStage class was pretty good, but I learned far more actually getting my hands dirty and doing the conversions. The first 1 or 2 conversions probably were pretty slow, but after that the conversion process sped up significantly.
The great thing about rewriting existing programs is that you have a baseline to compare to - a) you know how long it takes for the original to run, so you can gather some performance-gain statistics, and b) you know the correct results for the original so if your converted program duplicates the results then your testing is done. The other benefit is that you are not splitting your time/brain juice between designing a new process and trying to learn a new tool. You are just learning.
One last note - If you get to take a DataStage PX class, make sure you get an instructor that has done extensive work in PX. That was really my only complaint about the class I took a few years ago. At that time, we had about 25 people in our development group and ended up having to split up into 2 sessions of the weeklong class. The first class was taught by someone who used to work for Torrent and had done plenty of work in DataStage PX. The second class, the one I attended, was taught by someone who had never used Torrent and had done very little PX work. As a result, we probably knew more about parallelism (Torrent had been doing that all along) than she did. Now, to her credit, she did a great job teaching the tool and the PX version had only been out a little while so unless someone worked for Torrent and then DataStage, exposure to PX would be limited.
Anyway, that's my 2 cents on the whole conversion thing. I'm sure you really wanted to hear my long-winded ideas... :D
Brad.
We never used Server, but we had Torrent in house here. All of our Torrent code had to be rewritten as PX jobs. The DataStage class was pretty good, but I learned far more actually getting my hands dirty and doing the conversions. The first 1 or 2 conversions probably were pretty slow, but after that the conversion process sped up significantly.
The great thing about rewriting existing programs is that you have a baseline to compare to - a) you know how long it takes for the original to run, so you can gather some performance-gain statistics, and b) you know the correct results for the original so if your converted program duplicates the results then your testing is done. The other benefit is that you are not splitting your time/brain juice between designing a new process and trying to learn a new tool. You are just learning.
One last note - If you get to take a DataStage PX class, make sure you get an instructor that has done extensive work in PX. That was really my only complaint about the class I took a few years ago. At that time, we had about 25 people in our development group and ended up having to split up into 2 sessions of the weeklong class. The first class was taught by someone who used to work for Torrent and had done plenty of work in DataStage PX. The second class, the one I attended, was taught by someone who had never used Torrent and had done very little PX work. As a result, we probably knew more about parallelism (Torrent had been doing that all along) than she did. Now, to her credit, she did a great job teaching the tool and the PX version had only been out a little while so unless someone worked for Torrent and then DataStage, exposure to PX would be limited.
Anyway, that's my 2 cents on the whole conversion thing. I'm sure you really wanted to hear my long-winded ideas... :D
Brad.