The 20% extra server capacity can accommodate processing needs if a few servers fail, or if demand fluctuates. If one site goes down, you immediately have the average compute available. If the outage is prolonged, then you may need to run some triage and defer workloads or add capacity, which is getting easier to do in the virtual world.
But how likely is this to happen? EMC reports that a study shows of all the events that affect availability (including not only natural disasters but also business mergers, acquisitions and data center relocations), less than 1% requires a failover from one site to another. In other words, DR is a necessary but expensive insurance policy. By contrast, a CA approach provides the DR shield, allowing companies to put "insurance premiums" to work elsewhere while saving on the overall IT business continuity investment.
Note that the 20% redundancy offered in CA is much less than the redundancy in a standard HA + DR combination. Because of this, EMC claims its approach offers a potential 28% to 50% reduction in overall compute costs.
The CA model requires that applications be run in parallel (or failover in near-CA versions) between two geographically separated sites. The synchronization of workloads at two distributed sites has to occur with low enough latencies (less than or equal to 5 milliseconds) that no one will be able to discern any difference in performance.
EMC's VPLEX Metro enables users to have read and write access to the exact same information at the same time (for all practical purposes) in two separate locations. It can support that performance level at distances of up to roughly 100 km, and work with common third-party clustering products--notably, VMware, Oracle RAC, HACMP/PowerHA, MC/Service Guard, MSCS and VCS. In other words, IT doesn't necessarily have to change how it works to accommodate the applications in the EMC CA world.
But what about the distance factor? Typically, 100 km is not considered enough for true DR. Some organizations may be willing to take the risk of having a second site within the 100 km range, with the understanding that a major disaster (think Hurricane Katrina) may still bring down a CA architecture. Others may have to take additional steps, such as bringing in third-party site for worst case scenarios.
A good economic case can be made for a CA approach based on a reduced server count and power savings. But from an IT perspective, the simplification inherent in EMC's CA Services is a big benefit because maintaining both HA and DR environments is a significant challenge. Moreover, in CA, the IT personnel at the remote site now are not an afterthought but are fully involved in normal operations.
Still, IT organizations will have to perform thorough evaluations to justify a move to CA.
The consolidation of HA and DR into one CA system seems to be a logical evolutionary step. The same benefit (24x7 forever availability) at lower cost and greater IT simplification makes a CA approach attractive. However, it is also a major decision that IT organizations have to think through carefully. So even though technologies are available that can do the job, enterprises may still have to be convinced to go forward.
EMC is a client of David Hill and the Mesabi Group.