Press "Enter" to skip to content

Day: March 2, 2016

DBCC Extended Checks

The SQL Server CSS team has a new post on DBCC performance improvements:

Starting with SQL Server 2016, additional checks on filtered indexes, persisted computed columns, and UDT columns will not be run by default to avoid the expensive expression evaluation(s.)  This change greatly reduces the duration of CHECKDB against databases containing these objects.  However, the physical consistency checks of these objects is always completed.  Only when EXTENDED_LOGICAL_CHECKS option is specified will the expression evaluations be performed in addition to already present, logical checks (indexed view, XML indexes, and spatial indexes) as part of the EXTENDED_LOGICAL_CHECKS option.

 

For filtered indexes, CHECKDB has also been improved to skip records that do not qualify as being indexed by target NC index. 

This is telling me that we’ll want to have two separate CHECKDB processes, one which regularly runs CHECKDB (or CHECKTABLE) and one which occasionally runs CHECKDB with EXTENDED_LOGICAL_CHECKS.

Comments closed

Table Partitioning

Slava Murygin walks through partitioning:

Usually new Partition Ranges are added to the end of a Function/Schema, but in my example I’ll demonstrate how to split middle Partition.
At first we create new File Group. Then make it next within a schema and then split a Function:

All records, which are Less than 1/1/2015 and more than 1/1/2014, were moved to the new partition.

Partitioning is one of those topics I run into frequently enough to need to know it, but not frequently enough to memorize it; every time it feels like I’m starting from scratch.

Comments closed

Improving SMO Performance

Cody Konior shows how to use GetDefaultInitFields in SMO to improve performance, and also explains a bug when you try to do this:

It turns out has no default properties set, and the exception seems to occur after you try to set some. Interestingly though the way that boolean flag works is to set default properties on each SMO type; which means the solution is to turn it on for everything and then only disable it for this specific type; retaining any of the performance improvements throughout the rest of SMO.

Anything that makes SMO faster gets a vote of approval from me.

Comments closed

Using The YEAR Function

Kendra Little looks at the performance ramifications of using the YEAR() function:

But if we hover over that index seek, we can see in the tooltip that there’s a hidden predicate that is NOT a seek predicate. This is a hidden filter. And because this is SQL Server 2016, we can see “Number of Rows Read” — it had to read 9.3 million rows to count 1.9 million rows. It didn’t realize the 2006 rows were together– it checked all the females and examined the FakeBirthDateStamp column for each row.

Built-in functions can cause SARGability issues.

Comments closed

Switching To Identities

James Anderson shows how to do a table switch to switch a table without an identity column to one with an identity column:

The SWITCH statement can instantly ‘move’ data from one table to another table. It does this by updating some meta data, to say that the new table is now the owner of the data instead of the old table. This is very useful as there is no physical data movement to cause the stresses mentioned earlier. There are a lot of rules enforced by SQL Server before it will allow this to work. Essentially each table must have the same columns with the same data types and NULL settings, they need to be in the same file group and  the new table must be empty. See here for a more detailed look at these rules.

If you can take a downtime, this is pretty easy.  Otherwise, making sure that the two tables are in sync until the switchover occurs is a key problem to keep in mind.

Comments closed

System Database Restrictions

Dave Mason ran into a system database restriction regarding number of files:

One thing I recently learned is that you can’t add files to [model]. I was hoping to impact the number of files a database has when it is created. It’s common for db’s to be created, with just the standard mdf/ldf pair of files. Then the db is loaded with varying amounts of data. After a certain threshold, it makes more sense to have multiple data files, based on the environment. Sure, I could add files later and “balance” the data across the files. But it would be nice to *not* have to do this; some down time may be required and LOB data presents its own challenges

There aren’t too many restrictions on the model database, but this is a biggie.

Comments closed