Healthcare IT looks beyond high availability to continuous availability

Settling for traditional high availability, rather than a continuously available network paradigm, can be a costly strategy, eroding patient confidence and opening hospitals up to medicolegal issues, according to a SCAR University presentation. Turning to clustered computing could provide one solution to improving system availability.

"We need to try to move beyond mere disaster recovery," said Steve G. Langer, Ph.D., an associate professor of imaging physics and informatics at the Mayo Clinic in Rochester, MN.

Simply being prepared for a disaster and having the ability to recover lost patient files does not take into account the cost to patient convenience and confidence, not to mention hospital efficiency, that other types of downtime such as scheduled outages entail, Langer said.

He described two different types of downtime: unscheduled outages that occur due to natural disasters or network attacks, and scheduled outages that occur due to routine network system maintenance and upgrades by the vendor. Most vendors do not count the second type of downtime against a contracted guaranteed amount of uptime, he said.

When charting system uptime and availability, measurements are counted by the "nines" where:

• 99% uptime = 3.6 lost days a year

• 99.5% = 1.8 lost days

• 99.9% = 9 lost hours

• 99.99% = 1 lost hour

Citing the 2004 Meta Group, Langer said that scheduled downtime costs could amount to $142 per full-time employee per hour.

Hospitals should strive for four-nine (99.99%) availability and beyond, with three nines being the bare minimum, he said. They should try to work this into a contract with a vendor, consider a different vendor if their current vendor can't fulfill this requirement, or develop an appropriate charge for the actual downtime experienced.

In addition to scheduled downtime, networks, servers, and storage can be susceptible to failure. By moving to a clustered, or multimode, method of computing and away from single-node computing, hospitals are less likely to be stranded when one node fails.

Clustered computing can consist of dual-node active/passive clusters, dual-node active/active clusters, and load-balancing stateless clusters. Using a clustered computing methodology, hospitals can move away from a single point of failure bringing the entire facility down and toward a continuously available system.