Oracle High Availability Doug Smith CIS 764 Fall Semester 2007

Agenda ►Introduction ►Context ●System Quality Attributes ●Availability ►Oracle Real Application Clusters ►Oracle Data Guard ►Oracle Maximum Availability Architecture

System Quality Attributes ►Sometimes called “Non-functional Requirements.” ►Clements, Bass, and Kazman classification: ●Qualities of the system: availability, modifiability, performance, security, testability, usability. ●Business qualities: time to market, cost and benefit, projected system lifetime, legacy system integration, etc.. ●Qualities of the architecture itself: conceptual integrity, correctness and completeness, buildability, etc. ►Architect responsibilities: ●Elicit requirements related to system quality attributes. ●Provide an architecture that can realize the desired system quality attributes.

Availability Two ways to improve availability: ►Avoid system failure ►Repair quickly

Avoiding System Failure ►Common approach: redundancy ●Design system with sufficient redundancy to allow continued operation when components fail. ●Example: application tier ■Multiple load balanced servers ■Redundant load balancing appliances ■Redundant network switches, routers, etc. etc. ►Oracle approach: Oracle Real Application Clusters

Dealing with Disaster ►What if the data center is lost? ●Cantor Fitzgerald eSpeed, WTC North Tower, 9/11. ●Disaster recovery plan & mirrored data center ●Cantor back online when bond market reopened 47 hours later. ►Tools and processes needed to rapidly repair or recover systems to make them available again. ►Oracle tool: Oracle Data Guard

Oracle Real Application Clusters (RAC) ►Oracle RAC provides the ability to run an Oracle database across a set of clustered servers. ●Cluster strategy provides fault tolerance; a cluster server can be lost without impacting the functionality of the system. ●Provides a horizontal scalability capability; servers can be added to support additional users and load

RAC Physical Architecture Figure from “HP InfiniBand solution for Oracle RAC environments”

Application Partitioning ►Oracle marketing literature implies any application can be run unchanged on RAC and scale up by adding additional nodes to the cluster. ►However… ●Oracle RAC uses a “shared everything” clustering strategy. ●Contention between nodes for data blocks will limit scalability; content will become worse in some cases as nodes are added to the cluster. ●Application partitioning needed to reduce contention at the block level, and to reduce cross-instance data swapping. ►Oracle 10g provides work load management features ●Allows routing specific users or functions to specific nodes.

Oracle Data Guard ►Oracle Data Guard is a set of services for maintaining one or more standby databases. ●Transactionally consistent copies ●Both physical and logical copies are supported. ►Data Guard Services ●Redo transport ●Log apply ●Role Transition ■Standby to primary ■Primary to standby

Redo Logs: Context From “Oracle Architectural Components”

Oracle Data Guard Services Figure from “Oracle Data Guard Concepts and Administration, Oracle 10g Release 2”

Oracle Maximum Availability Architecture Figure from “Oracle Maximum Availability Architecture”

References Software Architecture in Practice, Second Edition by Len Bass, Paul Clements, Rick Kazman. Addison Wesley Professional, Cantor Fitzgerald - Forty-Seven Hours, Edward Cone and Sean Gallagher. Baseline Magazine, HP InfiniBand Solution for Oracle RAC Environments, C.pdf C.pdf Oracle Architectural Concepts, df df Oracle Data Guard Concepts and Administration, Oracle 10g Release 2, west.oracle.com/docs/cd/B19306_01/server.102/b14239/toc.htmhttp://download- west.oracle.com/docs/cd/B19306_01/server.102/b14239/toc.htm Oracle Maximum Availability Architecture,