Choosing the right database management system

Databases are the hidden workhorses of many organisations' IT systems, holding critical business intelligence and carrying out hundreds of thousands of transactions each day.

In many ways, the database has become a commodity. Products differ on price, performance, ease of database administration and functionality.

There is a huge choice of database management systems (DBMS), which includes packaged and open source database suites. The main suppliers include Fujitsu, Hewlett-Packard, Hitachi, IBM, Microsoft, NCR Teradata, Oracle, Progress, SAS Institute and Sybase.

Analyst firm Gartner has identified the main database suppliers in Europe as Oracle, with a 40.8% market share, IBM with 29.4%, and Microsoft with 14.9%. This gives the three companies more than 85% of the market, and all of them have enjoyed revenue growth over the past few years.

"Microsoft outpaced its two main rivals in terms of growth because of pent-up demand for SQL Server 2005, whereas Oracle had the largest incremental revenue increase year over year, growing revenue by more than £65.5m," said a Gartner report.

Oracle

The main Oracle database versions in use include Oracle 7, Oracle 8, Oracle 8i, Oracle 9i, and Oracle 10g, with Oracle Database 11g in beta. For Oracle, the database is a key part of its Fusion applications platform, although it is possible to use rival databases with Oracle's business software.

Several versions of the Oracle database are available, with different pricing and features to reflect how it may be used. The Standard Edition contains basic database functionality and is typically used on servers running between one and four processors.

However, users running the Oracle database on servers with more than four CPUs must convert to an Enterprise licence. Enterprise Edition has more features than the Standard Edition, particularly in the areas of performance and security. Enterprise Edition has no memory limits and can utilise clustering via Oracle Real Application Clusters software.

Also available are Express Edition, running on Windows and Linux Personal Edition, an enterprise version with a single usage licence and Database Lite, which runs on mobile devices.

Oracle 10g user Powergen implemented the database along with Oracle Warehouse Builder to centralise its customer information and analyse it to find out which customers were profitable and which were unprofitable.

Mark Perrett, customer relationship management manager at the utility firm, said, "The datawarehouse has become the centrepiece of our CRM infrastructure, allowing us to translate customer insight into actionable activity that directly improves our customer relationships."

IBM DB2

IBM's DB2 is the second most popular DBMS. IBM now refers to its DB2 database as a "data server" and, like the Oracle database, there are many flavours of the suite designed for a range of computers, from mainframes to handheld devices.

DB2 version 9, codenamed Viper, is the latest incarnation of IBM's DBMS. IBM offers several licensing arrangements that can allow users to avoid paying for database ­features they do not need.

DB2 versions include Workgroup, Workgroup Unlimited, and Enterprise Server Edition. The most sophisticated edition for Linux, Unix and Windows is DB2 Datawarehouse Enterprise Edition (DB2 DWE). This edition is designed for a mixed workload, such as online transaction processing with datawarehousing or business intelligence implementations.

DB2 DWE has several business intelligence features, such as extraction, transforming or loading, data mining, online analytical processing acceleration, and inline analytics.

Watch manufacturer Fossil Partners has used DB2 integrated into its SAP enterprise resource planning system for the past two years to make its global distribution network more efficient and add accessory products such as belts, handbags, sunglasses and jewellery.

Mark Reynolds, director for IT infrastructure and operations at Fossil, said he was planning to move to DB2 9, which has new features that make use of the latest MySAP suite.

Microsoft SQL Server

The third biggest selling database is Microsoft's SQL Server. Its growing popularity is partly down to its native integration with the Windows Server software stack, and also because of the technologies it uses, particularly in development, security and business intelligence.

There are four main versions of the latest edition, Microsoft SQL Server 2005: Express, Workgroup, Standard and Enterprise. Other versions include Developer, Mobile and Compact.

One major user of the SQL Server database is London Underground, which has integrated its main project management application, Primavera Enterprise 5.0, into a bespoke SQL Server 2000 database called the Master Project Database.

This software, which runs on a powerful Compaq Proliant DL740 datacentre server, handles 1,700 simultaneous projects for London Underground.

Open source alternatives

Alongside the big three database products, Gartner said systems from smaller suppliers have also grown in popularity, differentiating themselves by focusing on niche markets, such as embedded or mobile DBMS.

The Linux operating system is also gaining increased acceptance among suppliers and users as a database platform, according to Gartner, and many users are choosing it over Windows, Unix and mainframe database platforms.

"Open source DBMS products continue to improve in terms of functionality and scalability, and DBMS tool suppliers are beginning to provide support for these offerings," said Gartner.

Of the open source databases, Ingres, PostgreSQL and MySQL come out the best in Forrester Research's product evaluation, according to senior analyst Noel Yuhanna.

"Many enterprises are turning to open source databases to reduce database management costs and avoid supplier lock-in. The maturity of open source databases is at its highest level ever, with more choices, better support and comprehensive ecosystems," he said.

Open source databases from Derby, Firebird and Oracle are also strong performers, he added. "Ingres, Oracle and PostgreSQL offer strong support for transactional processing, while Oracle and MySQL offer strong support for embedded database platforms.

"For datawarehouses, none of the [open source] projects offer strong native datawarehouse-­related features, but some third-party suppliers help fill the gap with their extended offering for open source databases," said Yuhanna.

Another open source database rising in popularity is EnterpriseDB, which claims to be able to run Oracle-compatible applications at a lower cost. EnterpriseDB users include Sony Entertainment and Vonage, which have switched from commercial databases.

Measuring performance

Database performance is measured in several ways, and this can be useful in choosing the right product. The main benchmark is the TPC-C from non-profit body the Transaction Performance Processing Council (TPC), which measures online transaction processing performance.

The TPC benchmark aims to simulate real-world usage. A typical transaction would include updating a database system for such things as inventory control (goods), airline reservations (services), or banking (money).

In these environments, a number of customers or service representatives input and manage their transactions via a terminal or desktop computer connected to a database. The TPC would typically produce benchmarks that measure transaction processing and database performance in terms of how many transactions a particular system and database can perform per unit of time.

The current top 10 TPC-C performance results include servers running IBM DB2 9, Oracle Database 10g and Microsoft SQL Server 2005 databases. At the time of writing, the top score was an IBM System p5 595 server running DB2 9 producing 4,033,378 transactions per minute.

Other benchmarks include TPC-R and TPC-H for datawarehouses and decision support systems, and TPC-W for web-based database systems.

Supplier-based benchmarks from the likes of Oracle are also available, and there are several open source benchmarks, such as the Open Source Database Benchmark and PolePosition, an open source Java framework for benchmarking databases.

Benchmarks can help users to select the right database, and this is essential because, once chosen, few organisations will ever move off their platform, said Rob Hailstone, software infrastructure research director at Butler Group.

"It is a pretty fraught thing to change a database. Once you have an established system, the last thing you want to do is destabilise it," he said.

For users who do switch database platforms, migration issues can come from individual business applications working in certain ways with particular databases databases requiring certain back-up and recovery processes or having different ways of failing over.

"Databases are meant to be inter­operable, but they all have their own operational procedures and processes for storing data and so on. It should be a last ditch choice to change your database, although there are sometimes good financial reasons for changing, such as licensing issues, or acquisition and mergers," said Hailstone.

The way forward for databases is to have a virtualised master data management layer, which can feed requests into a pool of databases, said Hailstone. This means it is less important where the data resides, and it can be kept on several smaller databases arranged in a grid, which have in-built redundancy and failover capabilities.

Adding another layer of software will have a performance hit, warned Hailstone, but it could give smaller organisations a more cost effective alternative to running a larger, more expensive database and having to back up their data at a datacentre.

How to avoid datacentre downtime

More information

www.tpc.org

http://osdb.sourceforge.net