Oracle RAC – Beyond the Data Center
Over the years, Oracle’s Real Application Cluster (RAC) has become a mainstay of mission critical, highly available environments. Built as a shared-everything cluster, RAC offers an ideal solution for mission critical applications – due to its scalability, performance, and high reliability. That reliability is significant because RAC is designed to enable all operations to continue, regardless of underlying component failures. At the extreme, Oracle RAC can continue to operate even in the event of the loss of key components such as one or more cluster nodes, storage paths, storage devices, or IP networks.
With that as a background, it may seem strange to an outsider that the majority of Oracle RAC deployments today employ a DR strategy utilizing host-based and/or array-based replication, and a re-start of the RAC environment in the event of the loss of the primary site. This sort of recovery strategy is disruptive to users and applications, yet it has become the industry norm.
In contrast, few organizations implement Oracle RAC environments where the cluster nodes are distributed across multiple disparate sites. Fully supported by Oracle, the administrative and infrastructure complexities associated with host-based mirroring to stretch RAC over distance have indeed been formidable for the mainstream user community. The challenge most often cited is the set-up and ongoing administration and maintenance of host-based mirroring.
With Oracle’s recent certification of EMC VPLEX Metro for RAC, the door has now been opened for users to build continuously available RAC infrastructures that stretch far beyond the primary data center. Now, users can build non-stop business solutions that remain up and running – without the loss of a single transaction – even if the primary data center goes completely offline.
This is all possible because VPLEX Metro presents a cache-coherent copy of data at two sites that can be accessed simultaneously at both locations. In traditional array-based replication, the primary site is considered the active site, and the DR site operates as a passive site (where read/write access is not permitted). VPLEX Metro enables both sites to be active, in a structure EMC refers to as active-active. An easy way to envision active-active in practice is to consider a single array being physically stretched over distance – everything possible with that array in a single data center is now possible with that stretched array over distance.
With extended Oracle RAC, failure of the primary site is no different than the loss of one or more nodes in the case of a single site implementation. All data for the cluster remains online, and available at the secondary site – nodes at the surviving site continue the cluster’s processing without interruption. Since extended Oracle RAC with VPLEX Metro is implemented no differently than single-site RAC, the VPLEX solution requires no additional software or special configuration within the Oracle RAC infrastructure.
Administration of Oracle RAC stretched with VPLEX Metro is no different than administration of a single-site implementation. The traditional hardware RAID functions provided by enterprise-class storage arrays can now be extended over distance using distributed hardware RAID for extended Oracle RAC with VPLEX Metro. With VPLEX Metro, RAC administrators don’t have to be concerned with items inherent in host-based mirroring technologies, like path preferences, failure groups, load balancing and rebalancing as the infrastructure scales, or myriad other complexities. In essence, VPLEX removes storage administration tasks from the DBA so that they can focus on database administration.
The advantages of stretching Oracle RAC over distance with VPLEX are numerous. To start with, users benefit from a non-stop operational version of RAC – no longer is the possibility of downtime associated with site failure a concern. In addition, asset utilization is improved, because infrastructure that used to be on standby at the DR site is now actively in operation. Traditional DR planning is greatly reduced – with VPLEX, both sites are already continuously processing information. No longer is there a question whether the site will “come online” - both sites are continuously available.
From a networking perspective, Oracle RAC with VPLEX is far simpler and much less difficult to build. With host-based replication, users have to create complex, merged SAN infrastructures that span distance – which can introduce numerous failure scenarios. VPLEX does not require a merged fabric to span the two sites, instead the VPLEX infrastructure shares information directly over FC or IP.
Today’s solution for extended Oracle RAC based on VPLEX Metro is a powerful new option for users with mission critical continuous uptime requirements. By leveraging the unique functionality in the VPLEX Metro active-active technologies and the core capabilities inherent in Oracle RAC, the evolution of disaster-tolerant uptime has reached another milestone.