10.4 Maximum availability architecture

---

In the previous sections we discussed the various system availability requirements and how some of these requirements are met with the standard out of the box implementation of RAC. This level of availability is normally sufficient for most organizations operating on a 24*7 schedule providing support for their day-to-day customers. However, there are organizations that have machine-critical, high-transaction processing, state- of-the-art technology databases that are required to be kept up all the time; to mention a few would be the stock exchange, the space station, etc, where every second of downtime is very expensive. These organizations require something called a maximum availability architecture (MAA).

MAA provides a redundant and robust architecture that prevents different outages or recovers from an outage within a small MTTR. The goal in this situation is to have no impact or minimal impact on availability while catastrophic outages can be repaired.

When Internet based business requirements stipulate an uptime of 99.999%, the hardware infrastructure is made redundant at all tiers, such as the network tier, application server tier, load balancing tier, etc. On the database tier, the clustered database technologies such as RAC provide continuous availability by having one instance up when one fails.

This level of redundancy is available or is concentrated at one location (primary data center) of the enterprise. What if a catastrophic outage (such as an act of god) caused the entire data center to collapse? The entire business operation is affected. The redundancy at the local data center does not provide much room for availability. From our example above, the stock exchange or the space station would be seriously affected if such a situation arose. This calls for a much higher level of redundancy.

MAA is obtained by providing redundancy to the already redundant architecture, that is the entire primary data center is made redundant by creating an identical operation infrastructure at a remote location. If the data center had two application servers, two database servers, two load balancers, two firewalls, the remote redundant site will also have the exact same set of hardware infrastructure. Fully functional and ready to be active when any catastrophic outage happens.

Changes to most of these tiers are less frequent compared to the database tier, where the data is consistently changing. In this case data needs to be moved to the remote location as close to real time as possible, that is data needs to be moved from the primary RAC database to a remote RAC database. This could be implemented using Oracle utilities such as ODG or OAR.

---