DSXchange

To address the co-existence issue, consider all the ways that ODI could be illustrated within Information Server for lineage and governance, using the OpenIGC API ...where you can create your own classes of objects, and connect them as desired for lineage -- even for "deep down" internals of other ETL tooling, if you desire to take it to that level. There have been some people exploring the parsing and import of ODI metadata for that purpose. CWM XMI isn't going to cut it --- except in rare circumstances, it is mostly a lowest common denominator type of metadata representation --- but there are probably lots of ways that ODI metadata can be exported, parsed, and put into a format that would be useful for an OpenIGC scenario.

Other than that, it is certainly well documented where ODI excels and where DataStage excels --- the greater the variety of your sources and targets, the greater the need for tooling such as DataStage and other non-DBMS-centric transformation. If you are 100% Oracle, then the decision may be simpler.

Ernie

I got the impression from the IBM web site that DataStage Balanced Optimization features are included in version 11.5. Previously it was an additional cost option.

My biggest question right now is - where is part 1? A (mostly) honest question as I figured it must have been something I didn't recall reading but a search failed to turn any such thing up. Hence the question.

rameshrr3 wrote:1. Do you find that if you follow the recommended patterns to push as much processing to the database, Balanced Optimization will be able to result in a 100% ELT outcome?

Yes - it does a pretty good job of realizing when it can dump everything into the database and creating the appropriate SQL to to do so.

rameshrr3 wrote:2. What are some real-world results using it? Do you find that only ideal cases are 100% converted leaving the majority to be mixed or 100% DataStage?

We were very judicious in our usage. We were in an environment with lots of databases, so many of our jobs were "mixed" because they weren't Oracle to Oracle. Also, you are assuming that the Oracle system is big enough (and has bandwidth) to out-perform the DataStage server, which may or may not be true. We did have several jobs that were not pushed down to Oracle when testing showed it would be slower because the Oracle server was overwhelmed during that time-frame. It was faster to offload the processing to DataStage and then post the results back to the Oracle server.

rameshrr3 wrote:5. What do you think are its major strengths (ideal cases) and weakness (non-ideal cases)? Are there any areas where DataStage Balanced Optimizer shines compared to ODI?

All I can do is re-iterate much of what Ernie said earlier. Mixed database environments, or environments with limited SQL skills. Several times I've been paid to take pages of complex SQL and re-write it in DataStage because the existing staff couldn't support the SQL job. They were willing to take the performance hit to get something they could easily modify and extend. Now at 11.5 (with Balanced Optimization included), they can still use the GUI but push the processing back down to Oracle for improved performance. Best of both worlds.

Sorry - don't remember enough details to address other two questions.

@Chulett : (Part 1) is on the same shelf where windows 9 and datastage 10 are

Part1 is just what I consider previous debates on this topic , but did not want to hijack those older threads where I was not the OP.

rameshrr3 wrote:@Chullet : (Part 1) is on the same shelf where windows 9 and datastage 10 are

DataStage 10 does exist - it's the engine underneath IBM's Master Data Management server. It was never released as an independent product, however, for reasons that are internal to IBM Marketing.

Does the internal version of datastage engine have anything to do with the versions of the Universe Engine being release by Rocket UV ? Because the currently marketed version of UV is 11.x. I had an impression the 2 went their own way after UV version 9.

Rocket U2 UniVerse and DataStage engine did indeed go their own ways (although many similarities remain, the development trajectories have been quite different) when IBM took over UniVerse from Informix in 2000/2001. That the version numbers are currently both 11 is purely coincidental.