OS for DataStage 8
Moderators: chulett, rschirm, roy
OS for DataStage 8
I am new to DataStage and have been tasked to set up a IIS/DataStage 8 Enterprise Edition environment on an IBM HS21 Blade Server with 4 dual core xeon cpu's. We will be using Oracle 10g r2 for the metadata server that is running on Linux. What are the pro's and con's of using Windows Server 2003 rather than Linux (RHEL4) for DS8?
The differences overall between the two OSs for the repository are not that significant and chances are high that you won't notice any differences from DataStage.
This means the decision is not truly technical but related to your infrastructure. Do you have more OS level expertise on one platform or the other? Does your DBA hate one system or the other? Do you already have licenses?
This means the decision is not truly technical but related to your infrastructure. Do you have more OS level expertise on one platform or the other? Does your DBA hate one system or the other? Do you already have licenses?
</a>I will am more comfortable with Windows but I recall someone saying that Enterprise works better on Linux than on Windows. Our current environment is DS6 running on SUN but I have not had much involvement with it from the admin side. Since we will be rewriting many of our jobs I just want the best platform from the performance and reliability standpoint. We have an 8 CPU license and our IBM rep told me to let them know which platform we build.
The repository database is not heavily used (compared to a DWH or staging DB) so any performance differences should be negligible.
More important is your choice of platform for the actual DS Engine and for the DWH database or target - their configuration and sizing will have a much more significant impact than your choice of repository; to use a car analogy the repository is like your car's battery - it needs to be there and gets used, but doesn't impact your vehicle's top end speed. The DS Engine and DB Engine would then be like your car's engine and drivetrain.
More important is your choice of platform for the actual DS Engine and for the DWH database or target - their configuration and sizing will have a much more significant impact than your choice of repository; to use a car analogy the repository is like your car's battery - it needs to be there and gets used, but doesn't impact your vehicle's top end speed. The DS Engine and DB Engine would then be like your car's engine and drivetrain.
I guess I was not clear 
- I was asking about the platform for the DS Engine and Domain installation. Either Windows 2003 or Linux RHEL 4 are supported in our data center.
The following is the known set up for Database and won't be changed based on the OS for the DS Engine:
Metadata Repository on Oracle 10g
Database for resultant data will be Oracle 10g (DB Engine?)
Part of the driver for this question is that I have setup a test installation of DS8 on windows but haven't had the same success on Linux.
- I was asking about the platform for the DS Engine and Domain installation. Either Windows 2003 or Linux RHEL 4 are supported in our data center. The following is the known set up for Database and won't be changed based on the OS for the DS Engine:
Metadata Repository on Oracle 10g
Database for resultant data will be Oracle 10g (DB Engine?)
Part of the driver for this question is that I have setup a test installation of DS8 on windows but haven't had the same success on Linux.
I guess I misunderstood the question as well.
So you are looking at running either Windows 2003 or Linux on the IBM Blade Server that will host the DS Engine - same HW but different OS's?
I always envision Windows to use a lot more of the system resources than UNIX and be less efficient with process exchange/scheduling; but I gather that a lot of advances have been made by Microsoft over the years. Process exchange is much more important with PX than with Server installations, as PX tends to flood the machine with lots of processes and let the OS take care of efficient scheduling.
If you think that you won't be maxing out on CPU but will be primarily throttled by I/O then I would guess that the OS doesn't make too much of a difference.
Does Windows 2003 have upper addressable memory limits (like the 3Gb on other implementation)? If yes, then LINUX is the way to go.
So you are looking at running either Windows 2003 or Linux on the IBM Blade Server that will host the DS Engine - same HW but different OS's?
I always envision Windows to use a lot more of the system resources than UNIX and be less efficient with process exchange/scheduling; but I gather that a lot of advances have been made by Microsoft over the years. Process exchange is much more important with PX than with Server installations, as PX tends to flood the machine with lots of processes and let the OS take care of efficient scheduling.
If you think that you won't be maxing out on CPU but will be primarily throttled by I/O then I would guess that the OS doesn't make too much of a difference.
Does Windows 2003 have upper addressable memory limits (like the 3Gb on other implementation)? If yes, then LINUX is the way to go.
We ran on Windows originally when 7.5x came out. It ran OK, but I did run into some issues that were not easily resolved. One issue involved a transformer compiler bug with a parallel transformer that had alot of columns (500 or so). I think this is fixed now and I'm sure the code is more robust then when the first port to Windows was released. We however decided to move to Linux and have never looked back. That would be my suggestion, but again it's all up to what your SA's are comfortable supporting.
Good Luck!
Good Luck!
I think that there is a type of generation gap when it comes to using Windows as the OS for a heavy-duty commercial application; just as there remains a generation gap between the "heavy metal" mainframers and UNIX users. Just as UNIX was once considered to be a "toy" by mainstream data processing experts, the current crop of management and old farts (I include myself in one of those two groups, but have no direct reports) that have used UNIX for years often considers Windows to be barely efficient or stable enough to run a washing machine or touch-screen information booth. But Windows has now been around for years and in several incarnations and, although I am loathe to admit it, has "grown up"; perhaps the millions of FTEs (politically correct term for "Man Hours") spent on design, development and testing might have something to do with it.
So even though my initial reaction would be to go for a UNIX vs. a Windows solution, I know in my heart that I have no real rational reason for that bias and my newest notebook has neither a penguin nor Vista sticker attached to it.
So even though my initial reaction would be to go for a UNIX vs. a Windows solution, I know in my heart that I have no real rational reason for that bias and my newest notebook has neither a penguin nor Vista sticker attached to it.