"Sun Systems for Oracle Coherence"

The Sun BluePrints article "http://wikis.sun.com/display/BluePrints/Sun+Systems+for+Oracle+Coherence" "Sun Systems for Oracle Coherence"]" discusses a drop-in implementation of Oracle Coherence running on Sun servers utilizing Oracle Fusion Middleware components. Each major architectural component is described, providing information about the network interconnects, and detailing the functions of each tier in the solution. The paper also covers Oracle Coherence benchmark testing with a detailed review and analysis of the results.

Authors Nick Kloski, Nitin Ramannavar and Satish Vanga define the Oracle Coherence software platform as an in-memory data grid that provides numerous meta-services to help support and enhance the flow of data coming in from various sources. Oracle Coherence automatically and dynamically partitions data, optimizes the management of data based on the characteristics of information flow between applications and data sources, moves data closer to applications to reduce latency and improve performance, and offers a simple means to increase the capacity of shared data resources. It is these capabilities, the writers contend, that make the Oracle Coherence maximize Web-facing application performance and help foster near-linear scalability for many workloads. It can also serve to off-load back-end resources, such as database servers and storage systems.

Benchmark test results presented in the article substantiate the scalability and performance of this Web-facing architecture.

"Sun hardware paired with the scalability and resiliency of the Oracle Fusion Middleware stack - consisting of the Oracle WebLogic Server and Oracle Coherence software - is an ideal match for online Enterprise and Web workloads," the authors conclude. "Testing of an online workload with Oracle Coherence on Sun systems shows a 40% decrease in database load. Creating an architecture that transfers the burden of scalability from the vertically-scaled database tier to the horizontally-scaled data grid can allow applications to reach new levels of scalability."

The benchmark also demonstrated:

up to 2.5 times faster eCommerce queries1 and transaction per second
applications can scale to a million operations per second
delivers near-linear scalability with superior Quality of Service
reduces CPU utilization in the database tier by up to 40% and supports up to 25% more users

Kloski, Ramannavar, and Vanga offer best practices derived during implementation and testing of the configuration. Matching the number of application servers, memory capacity, and quantity of processing nodes to the target workload characteristics is key in achieving the best performance. They state that organizations can take advantage of the pre-tested architecture, best practices, and performance results to help speed time to deployment and save on provisioning and implementation costs for similar or adaptable workloads.

A review of topics presented include:

Architecture Design Choices