Lukas Eder has a nice post explaining different forms of automatic join elimination:

We intended to fetch all customers and their addresses. But observe: We project only columns from the CUSTOMER table and we don’t have any predicates at all, specifically not predicates using the ADDRESS table. So, we’re completely ignoring any contributions from the ADDRESS table. We never really needed the JOIN in the first place!

And in fact, the optimiser can prove this too, because of the FOREIGN KEY constraint on C.ADDRESS_ID, which guarantees that every CUSTOMER record has exactly one corresponding ADDRESS record. The JOIN does not duplicate, nor remove any CUSTOMER rows, so it is unneeded and thus eliminated (by some, not all databases, will list each database at the end of the article).

So, the database can rewrite the SQL statement to the following, equivalent SQL statement in the presence of said FOREIGN KEY

Read on for a comparison across different products as well.

Bert Wagner looks at how parameter sniffing works, why it’s (mostly) a good thing, and how to gently explain to the server when it’s a bad thing:

All subsequent executions of that same query will go to the query cache to reuse that same initial query plan — this saves SQL Server time from having to regenerate a new query plan.

Note: A query with different values passed as parameters still counts as the “same query” in the eyes of SQL Server.

In the case of the examples above, the first time the query was executed was with the parameter for “Costa Rica”. Remember when I said this dataset was heavily skewed? Let’s look at some counts:

Check it out for a clear depiction of the topic.  One solution that Bert doesn’t have but I will use sometimes is to create local variables in a procedure and set their values equal to input parameters.  That way, the optimizer doesn’t have an assumption as to the value of the local variable.  But there are several ways to get around this when it’s an issue.

Kendra Little explains that SQL Server will cache parameter values for invalid statements:

The first time that dbo.ReviewFlags is executed after the database comes online, it’s with an invalid parameter, like this:

• EXEC dbo.ReviewFlags @Flag = null;
• GO

This is caught by the IF block, hits the RAISERROR, and goes down to the THROW block, and the output is:

• Msg 50000, Level 11, State 1, Procedure ReviewFlags, Line 8 [Batch Start Line 70]
• @Flag must be a value between 1 and 5

But even though SQL Server didn’t execute the SELECT statement, it still compiled it. And it also cached the plan.

Read on to understand the trouble this can cause, as well as a few ways of solving the problem.  This is a special case of parameter sniffing problems, but the solutions are the same as in the general case.

Dmitry Zaytsev explains the math behind why query plans can be so inefficient when dealing with a large number of joins:

Let’s talk about the sequence of table joins in detail. It is very important to understand that the possible number of table joins grows exponentially, not linearly. Fox example, there are only 2 possible methods to join 2 tables, and the number can reach 12 methods for 3 tables. Different join sequences can have different query cost, and SQL Server optimizer must select the most optimal method. But when the number of tables is high, it becomes a resource-intensive task. If SQL Server begins going over all possible variants, such query may never be executed. That is why, SQL Server never does it and always looks for a quite good plan, not the best plan. SQL Server always tries to reach compromise between execution time and plan quality.

There are ways you can help the optimizer, and one of my favorite query tuning books was all about table selection.

Kellyn Pot’Vin-Gorman shows an example of query hints in Oracle and in SQL Server:

Oracle hints were quite common during the infancy of the Oracle Cost Based Optimizer, (CBO).  It could be frustrating for a database administrator who was accustomed to the Rules Based Optimizer, (rules, people!  If there’s an index, use it!) to give up control of performance to a feature that simply wasn’t taking the shortest route to the results.  As time passed from Oracle 9i to 10g, we harnessed hints less, trusting the CBO and by Oracle 11g, it started to be frowned upon unless you had a very strong use case for hinting.  I was in the latter scenario, as my first Oracle 11g database environment required not just new data, but a new database weekly and a requirement for me to guarantee performance.  I knew pretty much every optimal plan for every SQL statement in the systems and it was my responsibility to make sure each new database chose the most optimal plan.  I had incorporated complex hints, (and then profiles as we upgraded…)

With the introduction of database version Oracle 12c, it became a sought after skill to use hints effectively again, as many new optimizer features, (often with the words “dynamic” or “automated” in them) started to impact performance beyond what was outside the allowable.

Read on for a nearly-equivalent query in the two database systems.