• Home
  • About Us
  • Contact Us
  • Privacy Policy
  • Special Offers
Business Intelligence Info
  • Business Intelligence
    • BI News and Info
    • Big Data
    • Mobile and Cloud
    • Self-Service BI
  • CRM
    • CRM News and Info
    • InfusionSoft
    • Microsoft Dynamics CRM
    • NetSuite
    • OnContact
    • Salesforce
    • Workbooks
  • Data Mining
    • Pentaho
    • Sisense
    • Tableau
    • TIBCO Spotfire
  • Data Warehousing
    • DWH News and Info
    • IBM DB2
    • Microsoft SQL Server
    • Oracle
    • Teradata
  • Predictive Analytics
    • FICO
    • KNIME
    • Mathematica
    • Matlab
    • Minitab
    • RapidMiner
    • Revolution
    • SAP
    • SAS/SPSS
  • Humor

Two Approaches to Scalable Database Design

December 23, 2018   Sisense

Any form of application used for data analysis is stringently dependant on its ability to retrieve queries fast. However, when working with larger or more complex datasets, as well as an increasing amount of concurrent users, the performance depends largely on the underlying analytical database – whether this is built into the application as part of a single-stack tool, or implemented via a separate data warehouse layer.

What Makes a Scalable Database?

Database scalability is a concept in analytics database design that emphasizes the capability of a database to handle growth in the amount of data and users. In the modern applications sphere, two types of workloads have emerged – namely analytical and transactional workloads. Planning for workload growth must take into account the operating system, database design and hardware design decisions.

The Traditional Approach: Denormalization

At the hardware level, vertical and horizontal scalability exist. Vertical scaling involves increasing the capacity of a single machine. The capacity of a single machine can be increased by adding more CPU cores, RAM and storage. Horizontal scaling involves adding more machines to cope with growing workloads. This article will only highlight database design decisions required for a scalable application.

At the database level configuration, schema design, indexing, and query design affect the capability of a database to scale. Each relational database system provides a set of parameters that can be tweaked to improve performance. Therefore it is advisable to check parameters that are used in the specific database system. Schema design offers the biggest opportunity for a system developer to make decisions that will affect the scalability and performance of a database. Transaction workloads require a normalized design while analytical workloads require a denormalized design. Each design will be briefly explained.

Redshift 770x250 770x250 Two Approaches to Scalable Database Design

In a transaction workload the main objective is to process a large number of short UPDATE, INSERT and DELETE queries. Such databases are normalized to the third normal form (3NF). In this form business entities are broken into small tables. The business entities to be used in informing how to break up the tables come from a thorough analysis of the business. Normalization reduces redundant storage of data and enables tables to be joined to others. Indexing in a transaction workload does improve query response but slows update, insert and delete queries.

Therefore indexing can speed up or slow down a transaction database depending on how it is implemented. An iterative indexing strategy of implementing an index and testing its effect is appropriate. Although partitioning was developed for data warehouses it can still be beneficial in a transaction system especially where databases are big. A 3NF assures data consistency and accuracy but performance may be reduced due to the multiple joins involved. Trying to optimize query performance in a transaction system will affect the core functions of recording business data. Therefore it is advisable to migrate data analysis to a specialized system that uses an analytical workload design.

In an analytical workload the objective is to process few complex queries that arise in data analysis. This design is optimized for fast query performance. The data is denormalized meaning the business entities that were broken into different tables in the transaction system are joined together into one table. In this design, the two major table types are dimension and fact tables. Dimension tables contain descriptive attributes about the business. Fact tables contain measurable attributes that show the performance of a business. For example, in a sales process, the dollar amounts are the facts while region, customer type, and product category are the dimensions. In this design indexing and partitioning are very beneficial in improving the response times.

When working with either workload query optimization is very important. One optimization technique is avoiding the use of SELECT * instead, you should list all columns. Loops and correlated sub-queries slow down performance so they need to be used judiciously.

ElastiCube: Improving Performance via In-Chip Optimization

The approach outlined above, while scalable, often means sacrificing a great deal of the granularity of the original data on account of the need to denormalize and index said data in advance. This also limits the ability of the typically less-technical, front-end users of analytical applications to perform ad-hoc analysis – as they can only access certain views or queries according to a predefined schema.

The ElastiCube present an alternative form of analytical repository designed for fast query retrieval and scale. It’s based on the notion that by employing In-Chip™ technology to maximize the capabilities of modern hardware, business intelligence tools can achieve optimal performance when processing analytical queries, even without sacrificing the granularity of the data or flexibility of the front-end analysis – since faster query retrieval eliminates the need for summary tables, indexes or denormalization.

Redshift 770x250 770x250 Two Approaches to Scalable Database Design

Tags: Data Analysis | Data Modeling

Let’s block ads! (Why?)

Blog – Sisense

Approaches, Database, Design, Scalable
  • Recent Posts

    • AI Weekly: Biden calls for $37 billion to address chip shortage
    • NOT WHAT THEY MEANT BY “BUILDING ON THE BACKS OF….”
    • Why Healthcare Needs New Data and Analytics Solutions Before the Next Pandemic
    • Siemens and IBM extend alliance to IoT for manufacturing
    • Kevin Hart Joins John Hamburg For New Netflix Comedy Film Titled ‘Me Time’
  • Categories

  • Archives

    • February 2021
    • January 2021
    • December 2020
    • November 2020
    • October 2020
    • September 2020
    • August 2020
    • July 2020
    • June 2020
    • May 2020
    • April 2020
    • March 2020
    • February 2020
    • January 2020
    • December 2019
    • November 2019
    • October 2019
    • September 2019
    • August 2019
    • July 2019
    • June 2019
    • May 2019
    • April 2019
    • March 2019
    • February 2019
    • January 2019
    • December 2018
    • November 2018
    • October 2018
    • September 2018
    • August 2018
    • July 2018
    • June 2018
    • May 2018
    • April 2018
    • March 2018
    • February 2018
    • January 2018
    • December 2017
    • November 2017
    • October 2017
    • September 2017
    • August 2017
    • July 2017
    • June 2017
    • May 2017
    • April 2017
    • March 2017
    • February 2017
    • January 2017
    • December 2016
    • November 2016
    • October 2016
    • September 2016
    • August 2016
    • July 2016
    • June 2016
    • May 2016
    • April 2016
    • March 2016
    • February 2016
    • January 2016
    • December 2015
    • November 2015
    • October 2015
    • September 2015
    • August 2015
    • July 2015
    • June 2015
    • May 2015
    • April 2015
    • March 2015
    • February 2015
    • January 2015
    • December 2014
    • November 2014
© 2021 Business Intelligence Info
Power BI Training | G Com Solutions Limited