Advice/Recommendation please? How to best handle big models with various schemas?

Question

Hi
Using DM 3.3.0.747.
We have several bespoke applications, each using its own data model in separate schemas.  They do share various entities/tables which are made available in a shared schema.  There are references to these shared tables, etc. (e.g. foreign keys)
We want to embark on a process to reverse engineer and then maintain these applications' schemas in Data Modeller.
I would really like your views on how to best approach this.  The main concern is that the bigger the design (or more models/sub-views) the slower actions like saving and loading it becomes.  Once you also have it version controlled more time is added onto saving and synchronising with the sub-version repository and between team members.
The thinking is to primarily have each application (and hence its own schema) in separate designs (projects), but include tables from the shared/core schema as needed.  Would this be a good/recommended approach?
How does one address potential changes made (differences introduced) that may conflict between these separate designs?
e.g. in physical model the schema becomes a user that has permissions, etc.  and that same user may be in multiple designs.  Each design is version controlled, but in the end only exists once in the database.  If each design is synch'ed with the database, DDL generated and executed, this can cause one design to add permissions, etc. that the second DDL from the second design removes or replaces with its own?
I hope my question/scenario is clear enough.  If not, please revert to me for more clarification.
Thank you & Regards

Answer 1

Hi Hans,
I think it's better to use DM 4.0 and have everything in one design.
1) You can import all schemas you need into one design (one relational model). Save that design - important to do it before using "remote objects" approach
2) Use "Create new models based on schema names" wizard available in context menu for relational model in browser. Wizard will create new relational model for each schema in originating model and create there tables and views that
belong to that schema. All objects in originating model become read-only "remote objects".
If table (in new model) has a foreign key that refer table in another schema (now that table is moved to another new model) then "remote object" presentation is created for referred table.
At the end you'll have one large model with read-only remote objects and set of relational models that keep the track of foreign keys between tables in different schemas in database.
You can keep the originating model (the large one) in order to have global picture or delete it.
Properties from physical model are also transferred to new models in DM 4.0.
3) Check "Show 'Select relational models' dialog" in "Preferences>Data Modeler" - upon "open design"  that will bring up dialog allowing to select which model   to be loaded. The choice is persisted and
applied next time you open design (no need to see that dialog again if no change in what to be loaded).
4) For each model you can create "DDL generation" configuration (one or more) and exclude remote object from generation in DDL. Permissions should be maintained in model where table is editable and where DDL for that table will be generated.
The thinking is to primarily have each application (and hence its own schema) in separate designs (projects), but include tables from the shared/core schema as needed.
I won't recommend it. I know it's possible to have remote object in another design but you'll face following problems:
1) You cannot use design level domains and definitions in data types model
2) Remote object from another design are sensible to location of that remote design. You need to maintain the same directory structure everywhere designs (all related) are checked out and they need all to be checked out before dependent design to be open
3) You need to add remote objects manually while wizard described above will do it for you in the scope of one design
The main concern is that the bigger the design (or more models/sub-views) the slower actions like saving and loading it becomes.
The number of models shouldn't be a problem with support of "Select relational models" dialog. Subviews will impact loading if only they are large and are visible when "save" is used.
Total number of objects in model doesn't impact "save' operation - only modified objects are saved.
Yes, there is overhead for versioned designs - one more reason to move to DM 4.0 - SVN client 1.7 is used there.
Philip