Database Management: Version Management (Part I: SQL Server Feature Bloat)

Database Management Series

In an old SSWUG editorial “SQL Server Feature Bloat” by Ben Taylor, the author raises the question on whether SQL Server features like the support for Unicode, the increase in page size for data storage to 8k or the storage of additional metadata and statistics create a feature bloat. He further asks on whether customers may consider other database solutions, and on whether this aspect is important for customers.

A software or feature bloat is the “process whereby successive versions of a computer program become perceptibly slower, use more memory, disk space or processing power, or have higher hardware requirements than the previous version - whilst making only dubious user-perceptible improvements or suffering from feature creep” (Wikipedia).

Taylor’s question seems to be entitled, especially when is considered the number of features added in the various releases of SQL Server. Independently on whether they attempt to improve performance, extend existing functionality or provide new functionality, many of these features target special usage and are hardly used by average applications that use SQL Server only for data storage. Often after upgrading to a new release, it may happen that the customers see no performance improvement in the features used or the performance even decays, while the new release needs more resources to perform the same tasks. This can make customers wonder on whether all these new features bring any benefit for them.

It’s easy to neglect the fact that the SQL Server is just used as storage layer in an architecture and more likely that some of the problems reside in the business or presentation layers. In addition, not always a solution is designed to take advantage of a database’s (latest) features. Besides, it may happen that the compatibility level is set to a lower value, so the latest functionality won’t be used at all.

Probably the customers hope that the magic will happen immediately after the upgrade. For some features like the ones regarding engine’s optimization are enabled by default and is expected a performance gain, however, to take advantage of the new features the existing applications need to be redesigned. With each new edition it’s important to look at the opportunities provided by the upgrades and analyze the performance benefit as there’s often a trade-off between benefit and effort on one side, respectively between technical advantages and disadvantages on the other.

The examples used by Taylor aren’t necessarily representative because they refer to changes made prior to SQL Server 2005 edition and there are good arguments for their implementation. The storage of additional metadata and statistics is neglectable in comparison with the size of the databases and the benefits, given that the database engine needs statistics so it can operate optimally. SQL Server moved from 2 KB pages to 8 KB pages between versions 6.5 and 7.0 probably because it offers a good performance with efficient use of space. The use of Unicode character set become a standard given that databases needed to support multiple languages.

Feature bloating is not a problem that concerns only SQL Server but also other database products like Oracle, DB2 or MySQL, and other types of software. Customers’ choice of using one vendor’s products over another is often a strategic decision in which the database is just a piece of a bigger architecture. In the TPC-H benchmarks SQL Server 2014 and 2016 scored during the last years higher than the competitors. It’s more likely that customers will move to SQL Server than vice-versa, when possible. Customers expect performance, stability and security and are willing to pay for them, as long the gain is visible.