OWB Architecture

Hi,
The following is the architecture which we have currently deployed.
OWB Repository for development on Windows 2000 server on a LAN.
OWB Repository for Production on RS6000-AIX machine based in SG using a 2Mbps link ( WAN ).
OWB Client on XP .
The developement is carried out using the development Repository which is in LAN and the response is very fast.
But when we export the mapping (.mdl) from our development env. and import it to the Production env., the response to our Production repository is very slow.
Similarly if we use OWB client and get connected to the Production repository to do any changes to the mapping, the response is very slow.
Can u suggest, what needs to be carried out to make the response fast.
Thanks
Mahesh

Hi,
1.) The production rep must be as near to the target as it can be.When the mappings are executed there is much traffic between the target and the rep.
2.) If you work with the design center on the production rep it also must be as near as it can be. On design time there is also much traffic between the Client and the rep.
3.) Another way is to increase the bandwith between client and rep and between rep and target.
4.) To reduce the traffic between target and rep you can also reduce the audit level.
Regards
Detlef

Similar Messages

OWB 9.0.4 for Linux

When will it be available ?
Suppose I have Design, Runtime & Target Schema in 9i Database installed on Linux, and Client on Windows.
Do I need whole Linux version of OWB only to install "OWB Server Side components" on database machine ?
In previous version there was no such requirement, and the OWB architecture was much more simple...

Posted this on Metalink already. Thought it might be good to crosspost here:
Setup: OWB repository on Oracle 9.2.0.2 on Linux, OWB client running on Windows XP.
When trying to install the runtime environment into the database on Linux (it's a test setup so want to use this one database for repostitory and runtime repository) I keep running into "INS0034: The Database Server needs to be configured by running Runtime Repository option locally. Please run the Runtime Assistant on the Server where the database is located."
I did run the serverside install on the windows box but as far as I can tell it only installs some stuff on the windows box in a seperate home. There does not seem to be any installation done of objects into the database. What am I missing ?

OWB Services

Gurus,
In the OWB 11g OWB architecture i understand that the database hosts the OWB repository and the Metadata. The Design Center and Repository browser are working from the client end. Do we need the Application server for running any of the OWB services.
Appreciate your thoughts.
Thanks,

Hi ,
Application server is not needed for OWB related services .
Sutirtha

USERS and access to OWB

Hi
I am a dba with no warehousing experience. I have been asked to configure and support an OWB installation v9.0.2.0.8. I have installed the server side runtime repository and target schema's - and have run the service_doctor.sql and evrything appears fine.
I have the following accounts now
1 x OWBRUNREP user (runtime repository schema)
1 x OWRUNACCES user (runtime access user)
1 x OWB9iDEV user (design repository user)
3 x target schema's (for data and deployment)
My question is - my users connecting through the client on their pc can only connect to the Design Rrepository user OWB9iDEV - is this normal ??? All other connections fail ???
I am assuming they have to connect to the design user --> then deploy into runtime environment using deployment manager for example ???
So is everything alright - or do I have a problem ??? I can;'t decide from the last 8 hours of reading docs ??
Any help greatly appreciated
[email protected]

Hi!
Yes, the first time Runtime Repository Assistant is run it needs to be run on the host where OWB Runtime is installed. That's because the Runtime Service starts a java process that runs on the OS and not in the database. Subsequent operations of the Runtime Repository Assistant (such as to add more target schemas) can be remote.
The association of the Target schema with a Runtime Repository happens when you create a Target Schema using the Runtime Repository Assistant, which is why this is the only way to create a target schema, i.e. you cannot just start using any arbitrary schema as target schema for OWB if it hasn't been created this way.
Here is a good resource for your and others future reference. The OWB Architecture Wite Paper http://www.oracle.com/technology/products/warehouse/pdf/Architecture%20_White_Paper.pdf provides a condensed version of the architecture information contained in the Installation Guide. I personally found it the easiest way to make sense of the architecture.
Nikolai Rochnik

Why we need Design Repository and Runtime repository seperately?

Hi All,
Anybody please tell me why we need design time repository and runtime repository seperately? And also I want to know the total OWB Architecture.
Thanks,
Siv
Edited by: Siv on Feb 10, 2010 10:40 PM

Hi Siv:
Concerning Repository:
Since OWB 10.2 there is only one Repository storing Design and Runtime-Components.
excerpt from http://download.oracle.com/docs/cd/E10926_01/doc/owb.101/b12146/over.htm (OWB 10.1.):
The Design Repository, installed in an Oracle database, stores the metadata definitions for all of the objects used in Warehouse Builder. This is where all of the design information is stored for the target systems you are creating. You can access metadata stored here using the client user-interface, or through OMB Plus, the Warehouse Builder scripting utility.
The runtime repository, installed in an Oracle database, stores all of the deployment data as well as data from executed mappings and process flows. This is where the target environment information is stored for the business intelligence systems you create. This includes connection information for all of the deployment locations.
Which version of OWB are you using?
Greetings
Guenther Herzog

OWB - Installation Architecture and Infrastructure configuration

I thought it is a good idea to throw this out to the OWB community to see what your thinking is.
This is my first project; we are currently on detail design. The Client requests a document that covers the following details.
Architecture and Infrastructure configuration for OWB - Oracle Warehouse Builder – which is to be targeted at an ehosting data centre.
•     What is the best practice either from Oracle
•     Is it best to go for a 2Tier or 3Tier architecture
Infrastructure being of interest is:
•     The use of Application server and web service for giving visibility and access to the runtime / design repository via the repository browser.
•     The applicable application service i.e.: Control Centre Service
Can this distributed over multiple servers if you application server differs from you database server.
•     Do you have any information similar to this you compiled in the past that I can use
•     Or is there any source you can direct me to get this kind of information

Can this distributed over multiple servers if you
application server differs from you database server.This is not an OWB question. This is a database question. You are asking something like 'can I run Oracle with Apache with Tomcat?'. Simple answer is yes. Compex answer is load balancing?
Cheers,
Donna

OWB bugs, missing functionality and the future of OWB

I'm working with OWB for some time now and there are a lot of rough edges to discover. Functionality and stability leave a lot to be desired. Here's a small and incomplete list of things that annoy me:
Some annoying OWB bugs (OWB 10g 10.1.0.2.0):
- The debugger doesn't display the output parameters of procedures called in pre-mapping processes (displays nothing, treats values as NULL). The mapping itself works fine though.
- When calling selfmade functions within an expression OWB precedes the function call with a constant "Functions." which prevents the function from being executed and results in an error message
- Occasionally OWB cannot open mappings and displays an error message (null pointer exception). In this case the mapping cannot be opened anymore.
- Occasionally when executing mappings OWB doesn't remember changes in mappings even when the changes were committed and deployed
- When using aggregators in mappings OWB scrambles the order of the output attributes
- The deployment of mappings sometimes doesn't work. After n retries it works without having changed anything in the mapping
- When recreating an external table directly after dropping the table OWB recreates the external table but always displays both an error message and a success message.
- In Key Lookups the screen always gets garbled when selecting an attribute as a join condition
- Usage of constants results in aborts in the debugger
- When you reconcile a table used in a key lookup the lookup condition sometimes changes. OWB seems to remember only the position of the lookup condition attribute but not the name.
- In the process of validating a mapping often changes in the mapping get lost and errors occur like 'Internal Errors' or 'Null Pointer Exceptions'.
- When you save the definition of external tables OWB always adds 2 whitespace columns to the beginning of all the lines following 'ORGANISATION EXTERNAL'. If you save a lot of external table definitions you get files with hundreds of leading whitespaces.
Poor or missing functionality:
- No logging on the level of single records possible. I'd like the possibility to see the status of each single record in each operator like using 'verbose data' in PowerCenter
- The order of the attributes cannot be changed. This really pisses me off expecially if operators like the aggregator scramble the order of attributes.
- No variables in expressions possible
- Almost unusable lookup functionality (no cascading lookups, no lookup overrides, no unconnected lookups, only equal condition in key lookups)
- No SQL overrides in soruces possible
- No mapplets, shared containers or any kind a reusable transformations
- No overview functionality for mappings. Often it's very hard to find a leftover operator in a big mapping.
- No copy function for attributes
- Printing functionality is completely useless
- No documentation functionality for mappings (reports)
- Debugger itself needs debugging
- It's very difficult to mark connections between attributes of different operations. It's almost impossible to mark a group of connections without marking connections you don't want to mark.
I really wonder which of the above bugs and mssing functionality 'Paris' will address. From what I read about 'Paris' not many if at all. If Oracle really wants to be a competitor (with regard to functionality) to Informatica, IBM/Ascential etc. they have a whole lot of work to do or purchase Informatica or another of the leading etl tool
vendors.
What do you think about OWB? Will it be a competitor for the leading etl tools or just a cheap database add on and become widely used like SAB BW not for reasons of technology or functionality but because it's cheap?
Looking forward to your opinions.
Jörg Menker

Thanks to you two for entertaining my thoughts so far. Let me respond to you latest comments.
Okay, lets not argue which one is better.. when a tool is there .. then there are some reasons to be there...But the points raised by Jorg and me are really very annoying. Overall I agree with both yours and Jorg's points (and I did not think it was an argument...merely sharing our observations with each other (;^)
The OWB tool is not as mature as Informatica. However, Informatica has no foothold in the database engine itself and as I mentioned earlier, is still "on the outside looking in..." The efficiency and power of set-based activity versus row-based activity is substantial.
Looking at it from another way lets take a look at Microstrategy as a way of observing a technical strategy for product development. Microstrategy focused on the internals (the engine) and developed it into the "heavy-lifting" tool in the industry. It did this primarily by leveraging the power of the backend...the database and the hosting server. For sheer brute force, it was champion of the day. It was less concerned with the pretty presentation and more concerned with getting the data out of the back-end so the user didn't have to sit there for a day and wait. Now they have begun to focus on the presentation part.
Likewise this seems to be the strategy that Oracle has used for OWB. It is designed around the database engine and leverages the power of the database to do its work. Informatica (probably because it needs to be all things to all people) has tended to view the technical offerings of the database engine as a secondary consideration in its architectural approach and has probably been forced to do so more now that Oracle has put themselves in direct competition with Informatica. To do otherwise would make their product too complex to maintain and more vendor-specific.
I am into the third data warehousing/data migration project and my previous two have been on Informatica (3 years on it).I respect your experience and your opinions...you are not a first timer. The tasks we have both had to solve and how we solved them with these tools are not necessarily the same. Could be similar in instances; could be quite different.
So the general tendency is to evaluate the tool and try to see how things that were needed to be done in my previous projects can be done with this tool. I am afraid to say .. I am still not sure how these can be implemented in OWB. The points raised by us are probably the fall out of this deficiency.One observation that I would make is that in my experience, calls to the procedural language in the database engine have tended to perform very poorly with Informatica. Informatica's scripting language is week. Therefore, if you do not have direct usability of a good, strong procedural language to tackle some complicated tasks, then you will be in a pickle when the solution is not well suited to a relational-based approach. Informatica wants you to do most things outside of the database (in the map primarily). It is how you implement the transformation logic. OWB is built entirely around the relational, procedural, and ETL components in the Oracle database engine. That is what the tool is all about.
If cost is the major factor for deciding a tool then OWB stands far ahead...Depends entirely on the client and the situation. I have implemented solutions for large companies and small companies. I don't use a table saw to cut cake and I don't use a pin knife to fall trees. Right tool for the right job.
...thats what most managers do .. without even looking how in turn by selecting such a tool they make the life tough for the developers.Been there many times. Few non-technical managers understand the process of tool evaluation and selection and the value a good process adds to the project. Nor do they understand the implications of making a bad choice (cost, productivity, maintainability).
The functionality of OWB stands way below Informatica.If you are primarily a GUI-based implementer that is true. However, I have often found that when I have been brought in to fix performance problems with Informatica implementations that the primary problem is usually with the way that the developer implemented it. Too often I have found that the developer understands how to implement logic in the GUI component (the Designer/Maps and Sessions) with a complete lack of understanding of how all this activity will impact load performance (they don't understand how the database engine works.) For example, a strong feature in Informatica is the ability to override the default SQL statement generated by Informatica. This was a smart design decision on Informatica's part. I have frequently had to go into the "code" and fix bad joins, split up complex operations, and rip out convoluted logic to get the maps to perform within a reasonable load window. Too often these developers are only viewing the problem through the "window" of the tool. They are not stepping back and look at the problem in the context of the overall architecture. In part Informatica forces them to do this. Another possible factor is they probably don't know better.
"One tool...one solution"
Microstrategy until recently had been suffering from that same condition of not allowing the developer to create the actual query). OWB engineers need to rethink their strategy on overriding the SQL.
The functionality of OWB stands way below Informatica.In some ways yes. If you do a head-to-head comparison of the GUI then yes. In other ways OWB is better (Informatica does not measure up when you compare it with all of the architectural features that the Oracle database engine offers). They need to fix the bugs and annoyances though.
.. but even the GUI of Informatica is better than OWB and gives the developer some satisfaction of working in it.Believe me I feel your pain. On the other hand, I have suffered from Informatica bugs. Ever do a port from one database eingine to another just to have it convert everything into multi-byte? Ever have it re-define your maps to parallel processing threads when you didn't ask it to?
Looking at the technical side of things I can give you one fine example ... there is no function in Oracle doing to_integer (to_number is there) but Informatica does that ... Hmm-m-m...sorry, I don't get the point.
The style of ETL approach of Informatica is far more appealing.I find it unnecessarily over-engineered.
OWB has two advantages : It is basically free of cost and it has a big brother in Oracle.
It is basically free of cost...When you are another "Microsoft", you can throw your weight around. The message for Informatica is "don't bite the hand that feeds you." Bad decisions at the top.
Regards,
Dan Phillips

Best Practice: Update OWB 10.2.0.1 to 10.2.0.3 ?

Hey OWB-Guys,
I've searched the forum and metalink, but I could not find what I'm looking for. I want to update my OWB from version 10.2.0.1 to 10.2.0.3 (I'm working on WinXP Pro Sp2, so I can't use 10.2.0.4, right?).
Our system architecture will change soon, so I have to start the Control Center Service locally on my client computer in the future. I think there will be no OWB installed on DB server, if I understood our DBA correctly.
What will I have to do, to update OWB? Install patchset from Metalink on my client computer? What about the repository owner and repository user in the database?
Do you have a document / link with a guideline through the things to do?
Thanks in advance and have a nice weekend!
Steffen

Hey, nobody who upgraded from 10.2.0.1 to 10.2.0.3 ??? Please help me with some general hints or links describing that issue...
Thank you!
Steffen

OWB 10.2.0.1 - integrate OWB with CA Unicenter AutoSys JM

We want to integrate OWB with CA Unicenter AutoSys JM as described in the oracle document (see link below)
http://www.oracle.com/technology/products/warehouse/pdf/OWB_AutoSys%20with%20updates%20mv.pdf
Our current setup is as follows:
We have created a bunch of mappings which we run manually daily using the run_task procedure.
RUN_TASK
( location IN VARCHAR2,
task_type IN VARCHAR2,
task_name IN VARCHAR2,
custom_params IN VARCHAR2 DEFAULT NULL,
system_params IN VARCHAR2 DEFAULT NULL,
oem_friendly IN NUMBER DEFAULT 0,
background IN NUMBER DEFAULT 0
RETURN NUMBER;
We intend to create process flows in OWB and then deploy them direct to CA Unicenter AutoSys JM.
Apparently, there is an AutoSys adapter for OWB see link:
http://supportconnectw.ca.com/public/enews/jobmgt/job042104.asp
but I cannot find any information about this anywhere.
Has anyone succesfully been able to schedule OWB jobs through Autosys?... and if so can you give a step by step guide on how this was achieved?
Thanks,
Olu
Edited by: oluakins on 15-Oct-2008 07:43

Hi,
since OWB 9i oracle changes the complete architecture so you cannot use old connectors. And if you look at this Link you see that there is no adaptor available yet. Only an oracle universal adaptor.
So I don't think you will get some experiences at all.
Regards,
Detlef

Strategy to create projects in OWB

Hi Fellows
We are starting using OWB and I like to know your expertise how to structure our projects in OWB
We want to use OWB for Data warehouse projects.
We have different layers in our architecture and each layer has its own schemas
for example, we have a data load layer, an integration layer, a start schema layer and a reporting layer.
Each have their own schemas.
We have development projects that many developers can work on the project.
My question comes when you do recommend to create project.
Do we create one project for all layers?
Do we create oner project for each layer(schema)?
Do we create project for every development project?
or any other criteria to consider to help us decide when we create projects.
Sincerely

Hey,
please create one project for all layers and for each layer one module.
For each development project create one collection.
All answers are valid if your development is made on site.
BR

What is the transport mechanism in OWB?

Dear all,
We will be using OWB to control the Transportable tablespace process from a "Reporting Mirror" of the E-Business Suite to a Data Warehouse.
The underlying partitions we will be moving are stored on ASM.
Am I correct in thinking that OWB, behind the scenes, will be using the DBMS_FILE_TRANSFER utility to copy the partitions from source to target?
(ftp does appear to be a clear option within OWB but we don't want to use this)
Many thanks,
Barry
Barry Andersen | Technical Architect | +44.7917.265217
Oracle Technical Architecture Consulting
Oracle Parkway, Thames Valley Park, Reading. RG6 1RA

You have an oldschool options aswell:
Connect as SYS DBA with CONNECT / AS SYSDBA command.
Shutdown the database instance with SHUTDOWN command.
Rename or/and move the datafiles at operating system level.
Start Oracle database in mount state with STARTUP MOUNT command.
Modify the name or location of datafiles in Oracle data dictionary using following command syntax:
ALTER DATABASE RENAME FILE ‘<fully qualified path to original data file name>’ TO ‘<new or original fully qualified path to new or original data file name>’;
Open Oracle database instance completely with ALTER DATABASE OPEN command.
If the datafiles that need to be changed or moved do not belong to SYSTEM tablespaces, and do not contain active rollback segments or temporary segments, there is another workaround that does not require database instance to be shutdown. Instead, only the particular tablespace that contains the date files is taken offline.
Login to SQLPlus.
Connect as SYS DBA with CONNECT / AS SYSDBA command.
Make offline the affected tablespace with ALTER TABLESPACE <tablespace name> OFFLINE; command.
Modify the name or location of datafiles in Oracle data dictionary using following command syntax:
ALTER TABLESPACE <tablespace name> RENAME DATAFILE ‘<fully qualified path to original data file name>’ TO ‘<new or original fully qualified path to new or original data file name>’;
Bring the tablespace online again with ALTER TABLESPACE alter tablespace <tablespace name> ONLINE; command.

Versioning, Deployment, Execution in OWB 9.2

For the warehouse created using OWB versions 9.0.3, my company used a versioning tool. All the PLS generated using OWB were moved to Source control system and from there deployed to Target Schema.
Now in 9.0.4 generated code contains location information that gets replaced by physical location only when code is deployed using OWB. So the old system of versioning and deployment doesnt work. We need to deploy using OWB only. In this case we loose all the versioning of mappings.
Question 1 :
Will you please suggest a solution how can we do versioning of our code now ? Versioning is very important for us. In case it is not possible then can we go back to old system of deployment and execution by doing some manual workarounds.
Question 2 :
Is it necessary to deploy thru OWB only because of Location factor. In that case if we design such that source and target tables are in the same then these mappings will not contain location information and we should be able to deploy them without using OWB ? What are the other disadvantages of using this approach.
Question 3 :
When executing mapping using the template provided by OWB, we need to provide location name and execute it from Runtime Repository owner. I guess, When I deploy outside OWB I will not be able to use this method of execution. Default parameters and custome parameters of the mapping are stored in run time repository under the location name at the time of deployment . This information is accessed by the template to execute mapping. ( is it true ??)
We are not going to use multiple physical configuration for the same mapping, so if we deploy not using owb, can we execute directly in Target schema, without using template. How can we provide the input parameter to Main function in this case ?
Question 4 :
Will we be able to get Audit information if we do not use OWB's recommended way of deployment and execution.
Vibha

Vibha,
Question 1:
What you can do, is store the entire deployment script using the deploy to file option and version control that file. The format will be an ASCII-based XML file that versions exactly the same as you would version the files that OWB generated in previous releases.
Also, you may be interested in the version control you can do inside of OWB. That is, as of 9.0.4 we enable you to create snapshots of your design, which you could compare and restore at any point in the future. In 9.0.4, we provided a beta UI on top of this feature (scripting would be the other way to go), but in 9.2 we provide a production quality UI. To enable the UI in 9.0.4, close the product, locate preference.properties in <owb home>\owb\bin\admin and change the setting for MCMUIEnabled from false to true. Please do consider migrating to OWB92...
Question 2:
The locations were also introduced to support multi deployment (e.g. development, test, production) with potentially different physical setups of the machines, but no design changes necessary. Also, execution uses the location concepts. Did you already read http://otn.oracle.com/products/warehouse/pdf/OWB_RuntimeChanges.pdf, which contains more information about the changes and why these are there?
Question 3:
To execute directly from the database, I recommend you download http://otn.oracle.com/sample_code/products/warehouse/files/run_my_owb_stuff.sql and use that as the basis for your execution. The entire architecture changed from 9.0.3 to 9.0.4... and the runtime platform now manages the execution. Via the script above you get a wrapper around the script that OWB provides out of the box (minimum version 9.0.4.10).
Question 4:
The generated code contains all the calls to log the audits, so in principle yes. Having said that, if you do not deploy through OWB... it will be difficult to get the correct references that OWB will use during its logging to be resolved. I hope that with the explanation above you will be able to deploy and run inside the OWB context and make the entire environment work for you.
Thanks,
Mark.

One schema for OWB Design repository, runtime repository and target schema

Currently we have contents of OWB Design schema, runtime schema and target schema all combined into one schema of the same database in OWB 9.0.2 as well as OWB3i. We like to move to OWB10g in very near future. Can we keep the same structure for convenience of migration in OWB10g? Is it mandatory that OWB design repository (and components) must be separate from OWB run time repository (and components) and target schema? In other words is it possible and workable to use only one schema to contain OWB design repository, OWB run time repository and target schema in OWB10g environment with repositories to be situated on Oracle v9.2.0.1.0? Also what special considerations should be taken to create the database v9.2.0.1.0 and installation of OWB10g. What are the problems/side-effects to have all in one schema?
Also please let me know how to install Oracle Workflow server to be used along with OWB. Will OWB10g work with repository on Oracle database v9.2.0.1.0?
Your prompt advice will be very well appreciated.
Sankar

The design repo is a metadata repo that stores all the design-time objects and so forth.
It is an architectural decision that you or your team need to decide on. There are many flexible ways to architect an OWB infrastructure.
Also, your repository users will be using the design repository on the the other DB instance to do their design work...potentially less people always hitting the target database all the time.
-Greg

Owb design repository

I am unable to log in OWB Design Browser for report generation on Design Repository. I am in a Oracle client machine as well as it is OWB client. It giving an error message, 'Invalid Oracle HTTP Server information' whenever I am trying to login OWB design browser.
I am providing ip address of my oracle database server i.e host machine, port no, user id. What could be the reason and what is the solution?
And also I want to know the exact data for all these three columns.

The design repo is a metadata repo that stores all the design-time objects and so forth.
It is an architectural decision that you or your team need to decide on. There are many flexible ways to architect an OWB infrastructure.
Also, your repository users will be using the design repository on the the other DB instance to do their design work...potentially less people always hitting the target database all the time.
-Greg

OWB repository

Hi all,
I'm doing a fresh installation of OWB. I have 2 servers and I want to install the OWB in one server and OWB Repository in the other.
There is any way to do this? I'm using 11g R2.
Thanks!

Yes, it is possible. See [Implementing a Remote Runtime|http://download.oracle.com/docs/cd/B28359_01/owb.111/b31280/install_rep02.htm#BABJDHCJ] in OWB Installation Guide (it seems that your case is "Control Center Service Installed on a Dedicated Computer")
But drawback of this architecture - you need to start Control Center service manualy from dedicated computer (from server with OWB software)
Regards,
Oleg

OWB Architecture

Similar Messages

Maybe you are looking for