Press "Enter" to skip to content

Category: Indexing

The Practical Costs of Index Fragmentation

Published 2022-03-29 by Kevin Feasel

Tibor Karaszi digs into index performance:

See numbers and diagrams at the end, or at the top. I measured a few cases: the difference between no external fragmentation and severe external fragmentation (over 99%). I have both a narrow index and a wide index, and I read one (1), 10,000 and 100,000 rows using index searches (“range scan”). There were obviously no difference reading 1 row so I exclude that from my discussion below. For the other cases the extra time with an extreme level of external fragmentation is (from lowest impact to highest) 7%, 10%, 13% and 32%. The highest number (32%) is when reading many rows from a narrow index, i.e. many rows per page. Again, this is with an extreme level of fragmentation.

What’s interesting is that for the most part, there’s a negligible difference between ~0% internal fragmentation and ~99% internal fragmentation. The follow-on question is, how much are defrag operations costing you in performance and when is the benefit worth the cost?

Comments closed

The Upsert Pattern and Unique Indexes in SQL Server

Published 2022-03-21 by Kevin Feasel

Michael J Swart recommends a unique index:

To avoid deadlocks when implementing the upsert pattern, make sure the index on the key column is unique. It’s not sufficient that all the values in that particular column happen to be unique. The index must be defined to be unique, otherwise concurrent queries can still produce deadlocks.

Read on for Michael’s thoughts as well as some smart commenters adding on.

Comments closed

Tuning Non-Clustered Indexes via Missing Index Suggestions

Published 2022-03-21 by Kevin Feasel

Kendra Little announces a new article:

We’ve just published a new article in the SQL docs, Tune nonclustered indexes with missing index suggestions . The article explains what the missing index feature is, limitations of the feature, and how to use missing index DMVs and missing index suggestions in Query Store to tune indexes.

Read on to understand the impetus behind this article and then go check it out yourself.

Comments closed

Understanding Plans: Seeks and Scans

Published 2022-03-03 by Kevin Feasel

Erik Darling made me rhyme. First up, data retrieval via seek:

People. People complain. People complain about SQL Server. That’s probably why I get paid to deal with it, but whatever.

One complaint I get to hear week in and week out is that SQL Server isn’t using someone’s index, or that there are too many index scans and they’re slow.

That might actually be a composite of like twelve complaints, but let’s not start counting.

Erik gives us three reasons why we might not see a seek. But wait, there’s more!

I’m not sure why scans got such a bad rap. Perhaps it’s a leftover from the Bad Old Days© when people worried about logical fragmentation and page splits.

What I mean to say is: scans are often treated with a level of revulsion and focus that distracts people from larger issues, and is often due to some malpractice on their part.

This is true–scans aren’t inherently bad and Erik gives us a better mental model to work with.

Comments closed

Index Design Guide Updates

Published 2022-02-24 by Kevin Feasel

Kendra Little has a guide for us:

We’ve recently updated the SQL Server and Azure SQL index architecture and design guide. This article is an in-depth guide to indexing in databases using the SQL Server engine, including SQL Server, Azure SQL Database, Azure SQL Managed Instance, and Azure Synapse Analytics.

Our recent update adds a table to categorize the types of indexes discussed in the article, clarifies B-trees vs B+ trees, and describes how row locators (aka “secret columns”) are used in nonclustered indexes.

Read on for more information and do check out the guide.

Comments closed

Filtered Indexes and Functions

Published 2022-02-16 by Kevin Feasel

Eitan Blumin looks at filtered indexes:

In fact, absolutely no functions of any kind can be used within the WHERE clause of a filtered index. Not even schema-bound user-defined scalar functions.

Unfortunately, as stated in the Microsoft Docs page about Filtered Indexes, the WHERE clause of a filtered index can only support simple comparison operators.

Well, it’s not entirely true, as you CAN actually use some functions, but on two conditions:

Read the whole thing. Eitan lays out one limitation of filtered indexes and provides a couple of potential workarounds.

Comments closed

Choosing between Duplicate Indexes

Published 2022-02-03 by Kevin Feasel

David Alcock wants to know what choice you make when all choices lead to the same conclusion:

As there is an index that has the LastName and FirstName columns the optimiser has opted for an index seek operator using the IX_Person_LastName_FirstName_MiddleName index, and if I look into the Plan XML I can see that it’s using a trivial plan: StatementOptmLevel=”TRIVIAL”.
This basically means there’s one obvious way to return the query results so the optimiser has avoided the cost of going through full optimisation and has elected to use this plan straightaway.
So what happens if I create an identical copy of that particular index, in fact let’s create five indexes that are exactly the same:

There’s a Mass Effect 3 joke in my intro line. But read on for the answer and also check out Barney Lawrence’s comment for a bit more elucidation.

Comments closed

Wide Non-Clustered Indexes as Kinda-Sorta Clustered Indexes

Published 2022-01-19 by Kevin Feasel

Grant Fritchey gets our hopes up:

Everyone knows that you only get a single clustered index, right? Wouldn’t it be great though if you could have two clustered indexes?

Well, you can. Sort of. Let’s talk about it.

Click through to see what Grant means. This is a thing that I’ve done occasionally, though much more often, I’ve ripped it out because Database Tuning Advisor suggested it and a credulous user took DTA at its word.

Comments closed