CBR - ORACLE GRID SERIES PART II

Despite its maturity, the market for database management systems continues to be highly competitive, with Oracle, IBM and Microsoft competing for the hearts and minds of both CIOs and database administrators.

Evidence of ongoing competition in the market comes from research firm IDC's take on the 2004 database revenue figures, which saw Oracle out in front with 43.1% of the $14.9bn worldwide relational database market, ahead of IBM with 30.6%, and Microsoft with 13.4%.

While market share figures are a consideration for IT purchasers, more important are the major vendors' product strategies. Here again there are differences between the top three that indicate that the 30-year-old relational database market continues to evolve rapidly to solve new business requirements.

For the last two years Oracle has chosen to differentiate itself through delivering high levels of service in terms of scalability, performance, resiliance and security through grid computing based on commodity hardware and storage.

Starting with the launch of the 10g product line at OracleWorld 2003, the company has been steadily adding new grid-enabled functionality to its Database product to take advantage of a drive towards virtualised environments across a grid of low-cost commodity hardware.

While many of the partitioning, virtualisation, scheduling and management technologies that enable grid computing are available at the physical infrastructure level, Oracle's strategy has been to integrate technologies at the data level that will enable users to exploit the potential benefits of grid computing with software, resulting in reduced TCO through higher utilisation of low cost hardware.

"We're building on capabilities that we've been putting into the technology for the last few years," says Dave Pearson, Oracle grid programme director, EMEA. Oracle's grid strategy builds on the company's Real Application Cluster (RAC) technology, which runs on SMP, as well as commodity servers including blades, and supports multiple databases and multiple applications, which are not dedicated to individual servers.

While RAC was first introduced with Oracle 9i, Oracle Database 10g brought dynamic provisioning of resources (storage, servers, and data) and policy based automated workload management. This allows mixed application workloads to be managed proactively in a single cluster.

Oracle's grid strategy exploits virtualisation features in hardware and storage with the additional benefit of extending the same benefits to software, data, and applications.

"The interesting thing about grid is that it's about delivering the highest level of service, not just through the virtualisation of physical resources, but also through the virtualisation of data," says Pearson.

According to Pearson, there are three ways in which companies are integrating information across an enterprise: through the consolidation of data types, the federation of data from multiple sources, and through information sharing.

Oracle Database 10g includes technologies to enable data to be integrated via all three approaches. Oracle Transportable Tablespaces enables data subsets to be moved quickly from one Oracle database to another, while Oracle Streams can stream live data between databases or nodes in a grid environment.

Meanwhile, Distributed SQL and Gateways enable the integration of data stored in non-Oracle databases and the integration of file-based data from third party applications, while many enterprises are increasingly using web services, also supported by 10g, to achieve real time data integration and federation.

Pearson cites a customer service centre where agents might need access to customer relationship, order processing, invoice, product data and supply chain applications, as an example of the requirement for the functionality.

"Financial institution call centres want to interact with customers in the context of the entire portfolio of products and the services they receive," says Pearson, adding that it should not matter what database or application that data is being generated from. "This is a good example of virtualisation because the location of the data and the technology it's held in are totally transparent to the user," he adds.

While infrastructure-level grid developments from server and operating system vendors, as well as data-level grid developments from Oracle and others, are making it possible to abstract enterprise software and data from hardware resources, there remain cultural issues that are holding organisations back from truly exploiting grid computing, according to Pearson.

"The silo infrastructures which characterise most data centres today are inherently inefficient. They are costly to run, deliver a poor return on investment and aren't designed to support the level of enterprise wide business processes and information sharing that companies need in today's world," he says. "These infrastructures have partly arisen because business units have been forced to put together a budgeted business case for meeting each requirement. The belief that business units own their own resources has led to server hugging."

In order to take advantage of new technical advances, businesses need to change this approach to server ownership. "Cultural attitudes are bigger barriers than technological barriers," Pearson says.

Another issue related to increased information integration is security - both internal and external - and Pearson admits that this is an area in which the grid advocates still have some work to do in order to satisfy the concerns of some potential customers because there is currently no trust model to support inter-enterprise grid computing.

"In a silo environment at least you have one application per server. Once you've decoupled the application from the server, security becomes more complex," he says, adding that the company's Grid Control product is designed to provide centralised security management, including single sign-on.

As well as directory-based authentication via Enterprise User Security, Oracle has also introduced Virtual Private Database and Oracle Label Security technology to overcome that complexity. Virtual Private Database provides fine-grain access control that enables per-user and per-customer data access within a single database by associating security policies with tables, views and table columns. Oracle Label Security gives administrators the ability to create and assign policies to rows and columns within an application.

Within 10g there have also been a number of advances in terms of automated management via features such as Oracle Database Automatic Memory Management and Automatic Storage Management, which takes advantage of storage technologies such as RAID and Logical Volume Managers to balance I/O from multiple databases across all the devices in a disk group.

Grid control also provides the centralised framework and interface through which administrators can proactively manage all Oracle database resources, from the online movement data using transportable tables down to tuning individual SQL statements to optimise performance.

"There is a perception that managing a hundred servers in a single cluster is more complicated than managing a hundred separate servers," says Pearson. "However, the ability to define policy-based services in 10g simplifies and automates the management of grid resources. Grid Control continually monitors the environment and uses the information it collects to take corrective action automatically whenever any required quality of service is at risk."

PART ONE OF THIS SERIES CAN BE FOUND HERE