Database Engine Tuning Wizard - DB_Owner vs. Sysadmin

Brian Orrell — Tue, 20 Jun 2006 09:03:00 GMT

So I discovered this today, and it did not seem obvious at all. Say you have a login who has dbo rights to a database, they cannot run the Database Engine Tuning Advisor (formerly known as the Index Tuning Wizard) until a sysadmin has run it at least once on the server. It's true, it works... but seems very odd.

In SQL Server 2000, only members of the sysadmin fixed server role could use the Index Tuning Wizard to tune databases. In SQL Server 2005, using Database Engine Tuning Advisor, members of the sysadmin role can still tune databases, but now users who are members of the db_owner fixed database role also can tune the databases that they own.

On first use, Database Engine Tuning Advisor must be started by a user with system administrator permissions to initialize the application. After initialization, both members of the sysadmin fixed server role and members of the db_owner fixed database role can use Database Engine Tuning Advisor to tune databases. Keep in mind however, that the db_owner role members can only tune databases that they own. For more information, see Initializing Database Engine Tuning Advisor.

Multi-Table Aggregate Queries - The Good, Bad and Ugly

Brian Orrell — Thu, 25 May 2006 13:02:00 GMT

A common scenario I have seen come up at clients is the need to report against a common granularity (say Customer) with multiple different aggregate calculations from multiple tables with varying one-to-many relationships. Here's a sample diagram to explain:

In this example, Orders and FrequentBuyerPoints are only related by their common parent. One customer might have five order records but only three frequent buyer records.

If we wanted to see a list of all orders, we would write something like this:

SELECT C.Name as CustomerName, O.OrderDate, O.Quantity, O.Price, P.Name as ProductName
FROM Customer C
JOIN [Order] O
on (C.Id = O.CustomerId)
JOIN Product P
on (O.ProductId = P.Id)
ORDER BY C.Name, O.OrderDate

And see something like this:

If we needed to see a list of all frequent buyer points, we would write something like this:

SELECT C.Name as CustomerName, FBP.AwardDate, FBP.Points
FROM Customer C
JOIN FrequentBuyerPoints FBP
on (C.Id = FBP.CustomerId)
ORDER BY C.Name, FBP.AwardDate

And see something like this:

If we needed the aggregate by customer for orders and by customer for points we would write two statements like this:

Select C.Id, C.Name, Sum(O.Price) OrdersTotal, Count(*) AS TotalRecords
FROM Customer C
JOIN [Order] O
on (C.Id = O.CustomerId)
GROUP BY C.Id, C.Name

Select C.Id, C.Name, Sum(FBP.Points) PointsTotal, Count(*) AS TotalRecords
FROM Customer C
JOIN FrequentBuyerPoints FBP
on (C.Id = FBP.CustomerId)
GROUP BY C.Id, C.Name

And see something like this:

Now... what if we need one report with both order information and frequent buyer point information by customer? A common pitfall would be to combine the two previous statements into something like this:

Select C.Id, C.Name, Sum(O.Price) OrdersTotal, Sum(FBP.Points) PointsTotal, Count(*) AS TotalRecords
FROM Customer C
JOIN [Order] O
on (C.Id = O.CustomerId)
JOIN FrequentBuyerPoints FBP
on (C.Id = FBP.CustomerId)
GROUP BY C.Id, C.Name

And this would be the result:

Not good. What is the problem? The problem is that the SQL is first performing the join between the three tables which results in a number of rows per customer that is the product of the number of rows of each of the sub-tables. For the customer “Brian Orrell”, he has 3 order records and 5 points records which results in 15 records when you join both. This causes the information for each row of one table to be repeated the same number of times that you have number of records in the other table. The aggregates therefore are typically much higher than they should be.

The answer to this is to use sub-queries and roll-up your data before you join the tables together. That way the three “tables” being joined each have one and only one customer record and corresponding aggregate value. The SQL would look like this:

SELECT C.Id, C.Name, OrdersTotal, TotalOrderRecords, PointsTotal, TotalPointsRecords
FROM Customer C

JOIN
(Select C.Id, Sum(O.Price) OrdersTotal, Count(*) AS TotalOrderRecords
FROM Customer C
JOIN [Order] O
on (C.Id = O.CustomerId)
GROUP BY C.Id) OrderAggregate ON (C.Id = OrderAggregate.Id)

JOIN
(Select C.Id, Sum(FBP.Points) PointsTotal, Count(*) AS TotalPointsRecords
FROM Customer C
JOIN FrequentBuyerPoints FBP ON (C.Id = FBP.CustomerId)

GROUP BY C.Id) PointAggregate ON (C.Id = PointAggregate.Id)

ORDER BY C.Name

And the result would be as expected: