Identifying and Fixing Forwarded Records Performance Issue

Before going through the Forwarded Records performance issue and resolving it, we need to review the structure of the SQL Server tables.

Table Structure Overview

In SQL Server, the fundamental unit of the data storage is the 8-KB Pages. Each page starts with a 96-byte header that stores the system information about that page. Then, the table rows will be stored on the data pages serially after the header. At the end of the page, the row offset table, that contains one entry for each row, will be stored opposite to the sequence of the rows in the page. This row offset entry shows how far the first byte of that row is located from the start of the page.

(more…)

Real-Time Operational Analytics and Non-Clustered Column Store Index

In this article, we will focus on real time operational analytics and how to apply this approach to an OLTP database. When we look at the traditional analytical model, we can see OLTP and analytic environments are separate structures. First of all, the traditional analytic model environments need to create ETL (Extract, Transform and Load) tasks. Because we need to transfer transactional data to the data warehouse. These types of architecture have some disadvantages. They are cost, complexity and data latency. In order to eliminate these disadvantages, we need a different approach.  (more…)

Using Indexes in SQL Server Memory-Optimized Tables

Introduction

In this article, we will discuss how different types of indexes in SQL Server memory-optimized tables affect performance. We will examine examples of how different index types can affect the performance of memory-optimized tables.

To make the topic discussion easier, we will make use of a rather large example. For the purposes of simplicity, this example will feature different replicas of a single table, against which we will run different queries. These replicas will use different indexes, or no indexes at all (except, of course, the primary keys – PKs).

Note, that the actual purpose of this article is not to compare performance between disk-based and memory-optimized tables in SQL Server per se. Its purpose is to examine how indexes affect performance in memory-optimized tables. However, in order to have a full picture of the experiments, timings are also provided for the corresponding disk-based table queries and the speedups are calculated using the most optimal configuration of disk-based tables as baselines.

(more…)

Example of Improving Query Performance with Indexes

In this article, we’ll look at how an index can improve the query performance.

Introduction 

Indexes in Oracle and other databases are objects that store references to data in other tables. They are used to improve the query performance, most often the SELECT statement.

They aren’t a “silver bullet” – they don’t always solve performance problems with SELECT statements. However, they can certainly help.

Let’s consider this on a particular example.

(more…)

Missing Indexes in MS SQL or Optimization in no Time

When executing a query, the SQL Server optimizer tries to find the best query plan based on existing indexes and available latest statistics for a reasonable time, of course, if this plan is not already stored in the server cache. If no, the query is executed according to this plan, and the plan is stored in the server cache. If the plan has already been built for this query, the query is executed according to the existing plan.

We are interested in the following issue:

During compilation of a query plan, when sorting possible indexes, if the server does not find the best index, the missing index is marked in the query plan, and the server keeps statistics on such indexes: how many times the server would use this index and how much this query would cost.

(more…)

SQL Server Index Backward Scan: Understanding, Tuning

Table indexing strategy is one of the most important performance tuning and optimization keys. In SQL Server, the indexes (both, clustered and nonclustered) are created using a B-tree structure, in which each page acts as a doubly linked list node, having an information about the previous and the next pages. This B-tree structure, called Forward Scan,  makes it easier to read the rows from the index by scanning or seeking its pages from the beginning to the end. Although the forward scan is the default and heavily known index scanning method, SQL Server provides us with the ability to scan the index rows within the B-tree structure from the end to the beginning. This ability is called the Backward Scan. In this article, we will see how this happens and what are the pros and cons of the Backward scanning method. (more…)

Query Optimization in PostgreSQL. EXPLAIN Basics – Part 2

In my previous article, we started to describe the basics of the EXPLAIN command and analyzed what happens in PostgreSQL when executing a query.

I am going to continue writing about the basics of EXPLAIN in PostgreSQL. The information is a short review of Understanding EXPLAIN by Guillaume Lelarge. I highly recommend reading the original since some information is missed out.

(more…)

Troubleshooting long running queries in MS SQL Server

Preface

There is an information system that I administer. The system consists of the following components:

1. MS SQL Server database
2. Server application
3. Client applications

These information systems are installed on several objects. The information system is used actively 24 hours a day by 2 to 20 users at once on each object. Therefore, you cannot perform routine maintenance all at once. So, I have to «spread» SQL Server index defragmentation throughout the day, rather than defragmenting all the necessary fragmented indexes at one stroke. This applies to other operations as well.

(more…)