[dba-SQLServer] Creating joins across multiple databases

Djabarov, Robert Robert.Djabarov at usaa.com
Thu Apr 1 12:51:21 CST 2004


Based on what I've seen so far, it's not a good idea to do cross-db
joins, and not only because of security issues as Francisco mentions,
but also because of questions like "Who the heck is this guy and what is
he doing in my swamp?" which immediately start popping up, which leads
to finger-pointing when there is a slow-down or outage.  It also
violates database designation (DSS, OLTP) when it's OLTP for a native
user while becoming a DSS for the new-comer.

BTW, Francisco, did you copy the article or retyped it?  If you copied
it, I see a minor problem with Author's syntax...


Robert Djabarov
SQL Server & UDB
Sr. SQL Server Administrator
Phone: (210)  913-3148
Pager: (210) 753-3148
9800 Fredericksburg Rd. San Antonio, TX  78288
www.usaa.com

-----Original Message-----
From: dba-sqlserver-bounces at databaseadvisors.com
[mailto:dba-sqlserver-bounces at databaseadvisors.com] On Behalf Of
Francisco H Tapia
Sent: Thursday, April 01, 2004 12:41 PM
To: dba-SQLServer at databaseadvisors.com
Subject: [dba-SQLServer] Creating joins across multiple databases

I'm including the latest article from Arthur Fuller on the Builder.com 
SQL Newsletter because I hope it can bring forward some positive 
discussion.  Plus I'm sure many of us have dealt with this particular 
situation or will deal with it in the future.

/<BUILDER.COM SQL Newsletter>/*
Creating joins across multiple databases*

Enterprise databases often comprise several databases--one from 
production, another from marketing, and so on. All of the databases rely

on a certain amount of common data, which is frequently copied to each 
database that needs it.

However, this copying routine creates a new problem: skew. Where there 
is duplication, there is at least the opportunity for skew. The solution

to this problem is often replication, or scheduled jobs that update each

copy of the common data from one "master" database. This approach works,

but sometimes a better method is to move all the common data into one 
database, and then access it as needed from the other databases. This 
eliminates both the skew and the replication jobs.

This approach typically requires joining some local tables to the common

tables. For example, suppose you have two databases: Common and Sales. 
The Customers table resides in the Common database, while the Orders 
table resides in the Sales database. You want to create a query that 
lists the Customer information along with the Order information. How do 
you join these two tables? The answer lies in prefixes.

SQL Server uses a dot-delimited nomenclature that extends outward to the

database and the server. To specify a column in a table in another 
database, name the other database:

SELECT CustomerID FROM Common.Customers

To join a table from the Common database to a table in the current 
(Sales) database, use this same nomenclature:

SELECT * FROM Common.Customers
INNER JOIN Orders
ON Common.Customers.CustomerID = Orders.CustomerID

Not every organization's application domain requires multiple databases.

If your organization does require multiple databases, it's easy to 
create joins across databases.

/Arthur Fuller has been developing database applications for more than 
20 years. He frequently works with Access ADPs, Microsoft SQL 2000, 
MySQL, and .NET./
/<BUILDER.COM SQL Newsletter>/

Yes it is easy enough to join accross multiple databases and even 
accross several Sql Server linked servers :)   (very cool)... however my

concern and one that was not addressed in this article is the security 
aspect of it... say you just have a user who calls the report... does 
the user need to use an application who's using an application role and 
thus gets better than normal priviladges on the server?  I have not been

successful in adding new users to another database w/o also having to 
takes some hard steps; my solution was to ... Create VIEWS on the target

database... add the user to that 2nd database, then give explicit selet 
crights on the views and then join accross from db2 to db1 via the sproc

using the views instead of the tables... so while the data is not 
duplicated, the user is, and now has rights to base views that represent

tables :( , so because I think I might be doing this wrong.. I'm hoping 
Arthur or someone else will chime in with a few clues :)

Thanks,

-- 
-Francisco


_______________________________________________
dba-SQLServer mailing list
dba-SQLServer at databaseadvisors.com
http://databaseadvisors.com/mailman/listinfo/dba-sqlserver
http://www.databaseadvisors.com




More information about the dba-SQLServer mailing list