Fragmentation Can Affect Execution Plans

Jonathan Kehayias explains how index fragmentation can potentially affect execution plans:

Index fragmentation removal and prevention has long been a part of normal database maintenance operations, not only in SQL Server, but across many platforms. Index fragmentation affects performance for a lot of reasons, and most people talk about the effects of random small blocks of I/O that can happen physically to disk based storage as something to be avoided. The general concern around index fragmentation is that it affects the performance of scans through limiting the size of read-ahead I/Os. It’s based on this limited understanding of the problems that index fragmentation cause that some people have begun circulating the idea that index fragmentation doesn’t matter with Solid State Storage devices (SSDs) and that you can just ignore index fragmentation going forward.

However, that is not the case for a number of reasons. This article will explain and demonstrate one of those reasons: that index fragmentation can adversely impact execution plan choice for queries. This occurs because index fragmentation generally leads to an index having more pages (these extra pages come from page split operations, as described in this post on this site), and so the use of that index is deemed to have a higher cost by SQL Server’s query optimizer.

Let’s look at an example.

Check out the example, but definitely read the comments as there are some good conversations in there.

Related Posts

Tuning Azure SQL Database

Tim Radney walks us through some of the tools we have available to tune Azure SQL Databases: Many instance-level items that you have been used to configuring on full installations are off limits. Some of these items include:– Setting min and max server memory– Enabling optimize for ad hoc workloads– Changing cost threshold for parallelism– […]

Read More

Finding Missing Index Hints in Query Store

Grant Fritchey shows us another place where we can find missing index hints: A couple of notes on the query. I cast the query_plan as xml so that I can use the XQuery to pull out the information. It is possible that the plan might be so large that you get an error because of […]

Read More


December 2017
« Nov Jan »