High availability is becoming more and more important for all systems in various sectors. Service and human relationships will be the key differentiating factor for future successful companies (we must be on-time with our promises and we have to remember that the client pays our monthly salaries.) This obsessive attention to customer service is a life-or-death issue in every business.
Extending Real Application Clusters is the natural next step to ensure high availability. RDC (Relational Database Consulting) has extended its Real Application Cluster expertise by completing two extended Real Application Cluster installations at a leading telecommunications company. RDC is a BEE success story, delivering excellent service to various public, international and blue chip corporate clients.
Extended Real Application Clusters, also referred to as stretched RAC or Campus clusters, provide a greater tolerance to local disasters such as flooding, fire or power outages at a data centre. In this architecture, all nodes at all sites are part of a single database cluster.
To date, the practical limitations for Real Application Clusters on Extended Distance Clusters stemmed from the high impact of latency. Fortunately, modern networking equipment makes it possible to manage the latency and deploy such systems across a business campus or even up to distances of 120km.
RDC's recent implementations were done between data centres situated on the same campus. The primary difference in the configuration of the extended distance cluster, other than the fact that the database nodes are situated in different data centres, is the manner in which the storage is managed. To maintain high availability, the data needs to be available at both data centres. This is achieved by using host-based mirroring, which is supported by Oracle (Oracle ASM failure groups are used to achieve this). Oracle also introduced an initialisation parameter, “ASM_PREFERRED_READ_FAILURE_GROUPS”, to enable the nodes to be configured to read from their local storage.
The two installations also make use of differing hardware platforms. The one uses IBM hardware with AIX, while the other installation was done using HP blades and Linux. The value of the extended Real Application Cluster has already been demonstrated when one of the data centres lost power due to human error during routine maintenance. Although the data centre and all equipment were down, the other half of the cluster in another data centre continued operating and servicing requests. The failed cluster nodes later seamlessly rejoined the cluster when power was restored to the data centre.
The installation was a great milestone achieved by RDC, especially with the shortage of skill in this specialised area.
Share
Editorial contacts