High Performance and Availability with Oracle RAC and PHP
Review real-world experiences in deploying a business-critical PHP app over Oracle Real Application Clusters.
By John Lim
Published June 2009
Running a software application that is able to work reliably through hardware and software failures is incredibly hard. Our client, a Malaysian bank, commissioned us to develop a Web-based financial application for them that would be used in an intranet setting and was considered mission-critical. This meant that they that wanted a system that was reliable and had automatic failover to another server whenever any component failed.
In this article, I will cover the network, architecture and design of our RAC application. Then I will discuss the real-world experiences and problems we experienced. The experiences in general are also relevant even if you are using Java or .NET or your application and not PHP.
Our Oracle RAC Setup
We proposed PHP for the application with the data stored in Oracle Database 10g Standard Edition using Oracle Real Application Clusters (RAC) on Red Hat Linux Enterprise 4. Our final system looks like this:
In our network, we have two load balancers, in active-passive mode to distribute HTTP queries to our two application servers. The application servers are running Squid 3, Apache 2.2, and PHP 5, and use Oracle’s client side load-balancing and failover to connect to the Oracle database cluster.Squid 3 acts as a front-end web server on port 80. Squid will cache and dispense to web browsers static files such as Javascript and CSS, and GIF images. When Squid detects a .php file is being requested, Squid passes the request to Apache listening on a different port; this ensures that Apache can concentrate on executing PHP scripts without distraction.
The Oracle RAC is a cluster of independent database servers that share the same data and cooperate as a single system. Our RAC is a two-node cluster using a multipath connection to a Hitachi Storage Area Network (SAN). We use cluster-aware file-systems on the SAN to ensure that we both RAC nodes are able to write to the same files simultaneously without corrupting the data. There is also a private interconnect between the two RAC nodes, which is just a 1GB private LAN for data synchronization purposes.
