THE SQL Server Blog Spot on the Web

Welcome to - The SQL Server blog spot on the web Sign in | |
in Search

Sergio Govoni

A place where share my experience on SQL Server and related technologies

Updateable Column Store Indexes in SQL Server 2014


Column store indexes had been released with SQL Server 2012 to optimize data warehouse workloads that have specific patterns, data are loaded through T-SQL scripts or through SSIS packages, several times a day, in any case the frequency is not important, only that the available data is loaded in the same execution. At the end of ETL process, data is read with reporting tools. Usually data is written one time, then read multiple times.

In SQL Server 2012 there was the non-clustered column store index only; like a traditional B-Tree non-clustered index, it was a secondary index. However, it differs from a traditional B-Tree index because it is based on a columnar structure, though, the base table remains organized by row (in a structure called row-store and saved in 8K data pages).

The column store indexes are part of Microsoft In-Memory Technologies because they use xVelocity engine for data compression optimization and its implementation is based on a columnar structure such as PowerPivot and SSAS Tabular. Data in column store indexes are organized by column, each memory page stores data from a single column, so each column can be accessed independently. This means that SQL Server Storage Engine will be able to fetch the only columns it needs. In addition, data is highly compressed, so more data will fit in memory and the I/O operations can greatly decrease.

Column store indexes structure

Before talking about new feature of column store indexes in SQL Server 2014, it is important to introduce three keywords: Segment, Row Group and Compression. In a column store index, a segment contains values for one column of a particular set of rows called row group. As it is possible to see in the following picture, each red and gray portions are segments. When you create a column store index, the rows in the table will be divided in groups and each row group contains about 1 million rows (the exact number of rows in each row group is 1,048,576; in other word there are 2^20 rows in each row group). Column store transforms the internal index organization from row organization to columnar organization and there will be one segment for each column and for each row group. Column store indexes are part of Microsoft In-Memory technologies in which data is compressed and the compression plays a very important role, so each segment is compressed and stored in a separate LOB.

This article does not detail the algorithms to compress data in column store index. At any rate, keep in mind that each segment stores the same type of data, so in a segment, there will be homogeneous data and in this scenario, the compression algorithms will be more efficient than the one used to compress table rows because row usually contains different type of data. In general, data compression can be implemented using different techniques such as:

  • Dictionary Encoding
  • Run-length Encoding
  • Bit Packing
  • Archival Compression (only in SQL Server 2014)
    • It can reduce disk space up to 27%

The techniques used by SQL Server to compress data are undocumented.

The following picture shows an example of row groups and segments in a column store index... continue reading full article here.


Thanks to my friend Viviana Viola for the grammar review.

Published Monday, June 16, 2014 10:04 PM by Sergio Govoni

Comment Notification

If you would like to receive an email when updates are made to this post, please register here

Subscribe to this post's comments using RSS


No Comments

Leave a Comment


This Blog


Privacy Statement