Database – One Database for Multiple Applications

application-designArchitecturedatabasedatabase-design

Has anyone worked in an environment where multiple in-house applications are all under one database with each application given its own schema within that single database? I don't just mean a few small, related applications in one DB, but like full on standalone applications being put into one DB with other standalone applications.

I'm dealing with an architecture like this that was put into place long before me and am trying to argue that a new standalone application should be put into its own separate DB, but I'm basically being told that this "put everything into one database" architecture makes more sense than a separate DB and I am kind of at a loss for what to even say. I feel like I can't find much info about this out on the internet because no one else is putting multiple applications into one database like this, but maybe I am completely wrong?

I am working with Microsoft SQL Server and .NET web applications.

Best Answer

Yes I have worked in a similar environment.

The difference between a schema and a database isn't a huge one from an application perspective. After all, all the databases on a server run on the same server. I wouldn't sweat it.

The key thing is you maintain a separation between the tables and not have one set reference another, or have queries that span domains.

I would leave the DB setup to the DBAs and just ensure my SQL was good.

Architecture

Does the system handle "any sport"? Is the idea that you put on your architecture astronaut hat, and build a generic system that can handle any future type of sport that may not even exist today?

If so, obviously having dynamically-named tables is a huge pain, so it would make sense to have a schema that supports n sports, if required.

That said, I have a very strong bias against this approach: this is almost always more work and leads to poorer results. Making a separate UI, schema, etc, for each sport will ultimately lead to better user experience and easier to maintain code, even if it does mean some superficial amount of duplication (how to avoid/minimize this is a separate question).

How do you handle players that play multiple sports? Do they get two entries (eg, you treat as different people) or are you trying to do something specific with them?

Use

So let's assume you don't do sports dynamically (eg, if someone wants to add a new sport, it requires development effort to add it).

Is there ever a time where you're displaying players (or any other object you mentioned) from more than one sport at a time?

I could see this for a search function, where you could search by player or team name (regardless of sport), but beyond that I can't imagine many use cases.

If you never need to do this, then your approach is perfectly fine. You can stop reading here.

Alternative Schemas

Views

I'm a fan of KISS. In 15+ years of software development, I continue to fall back to the philosophy "build the simplest thing that works".

So my initial reaction, assuming a cross-sport search function is really the only use case, is to create views:

SELECT PlayerName, 'NFL' as [Sport], TeamName FROM NFL_Players JOIN NFL_Teams ... 
UNION  
SELECT PlayerName, 'NHL' as [Sport], TeamName FROM NHL_Players JOIN NHL_Teams ... 
UNION ....

Of course, if you add a new sport, you have to add to the view. It may also be useful to include other common information but really that's dependent on what needs to get shown.

I'd try to keep all the sport-specific stuff in the View definition, so the search code doesn't need to have much or any specific code (besides maybe knowing how to link to /nhl/players/player-name vs /nfl/... or however your app does that).

Table Inheritence

Table inheritance can work, but is pretty complex. I don't have a ton of experience with it, and in fact, I think every time I've been involved in evaluating it we ended up doing something simpler (like I'm suggesting here).

So personally, I have yet to find why this would be useful, but maybe there is a convincing use case (that I don't know of) that justifies the complexity (eg, Table inheritance solves the use case better than any other solution).

Separate tables for sport-specific attributes

You could do a single players table which has attributes common to all players of all sports, and then another set of tables like nhl_players_details that contains a playerId and columns with additional info on the player. If there are a ton of common attributes, or you have many uses of "all players from all sports" then this may make sense.

Key value pairs for sport-specific attributes

Completely alternative approach: have a players table (again, with common attributes like name) and then a player_data table that has PlayerId, Sport, Attribute, Value. The attribute names entered would be sport-specific. This allows you to essentially add new attributes without modifying the schema (your code would still have to know to load/display them of course). The drawback is you lose some integrity: value would typically be a string field, so your app code would have to be resilient and handle potential failures converting the string value to a specific data type (like integer).

This concept can of course apply to Teams, Games, etc.

Web Development – Using One User Database for Multiple Websites and Applications

I would recommend separate databases for the applications and one database for the user accounts.

I would put any fine-grained access control within the individual applications.

I would also look to make the one 'user database for the user accounts' be accessed via an authentication service using industry standard protocols such as LDAP.