What's the biggest difference between a high-availability health care network and a high-availability business network?
No one dies when a high-availability business network goes down.
A little extreme, you think? Not really. Health care network availability has taken on more importance as network convergence has turned hospitals into dense networks of sophisticated, computerized medical gear. Take the network away and you’d be left with a lot of dumb devices that cannot perform the life-saving functions for which they were designed.
Because of the lifesaving nature of the networked devices, health care institutions need networks that are different from ordinary business networks on almost every level. Whether it's the basic power infrastructure, the core, the wireless portion or the wide area network (WAN), the network has to be designed to never go down, to accommodate fluctuating workloads and to meet a variety of other requirements peculiar to health care institutions.
This set of requirements for health care network availability has been changing since the late 1980s. That's when the term convergence emerged on the network scene to describe how vendors were combining multiple services though a single vendor or provider.
Usually, this meant the merging of telecommunications services with data communications services. Simply put, the typical network switch/router vendor could offer Voice over IP products, providing the convergence of voice and data over a single network infrastructure. It was a revolution. The typical business CIO found this proposal a smart way to eliminate dual wiring infrastructures. Why build separate infrastructures for voice and data when a single data infrastructure could do it all?
Convergence in health care networks had a much slower start than it did in business networks, but after 2000, health care networks started changing the core meaning of convergence. Convergence now is defined as a process where technologies with distinct purposes merge and overlap. Not only are health care networks embracing telecommunications and data communications convergence, but now biomedical devices are also merging and overlapping.
Examples abound. Radiology imaging services are now exclusively digital. Medical images of all modalities are being transferred over both LANs and WANs.
Remember the proverbial joke about a physician's handwriting only a pharmacist could interpret? Prescriptions are no longer exclusively handwritten by physicians. Special network prescription electronic services are the norm and the preferred method of pharmacies for filling orders because they ensure that the prescription is correct. And the services provide audit trails to legally protect both the physician and the pharmacy.
Nothing is more emblematic of the technology convergence in health care than seeing what has occurred within biomedical engineering departments. Four years ago, the biomedical engineering department at the health center where I work reported to the clinical services division. Last year, senior management decided that biomedical engineering would now be part of Information Services. The decision was driven by convergence.
Almost all the pieces of equipment managed and serviced by biomedical engineering departments have computers and network components within their core. Prior to 2000, most biomedical devices were standalone devices with no internal storage, management or communication capabilities. That has all changed.
All of these devices save lives. All of these devices need a network, and if that network fails people can die.
The challenge for a health care CIO, as opposed to a typical business support CIO, is providing a 24/7, high-availability network that can support the clinical requirements of the devices listed above. A high-availability business network always has maintenance windows of downtime. That would be a luxury for a high-availability health care network. So where does one start in designing a high-availability health care network? You begin at the very basic infrastructure foundation and then build up. This starts with electrical power and then the core design.
Consider health care network's electrical power needs
I am always surprised that many CIOs never review or factor electrical power into their network design. There are three electrical sources in a health care environment: street or raw power, emergency power and life-critical (LC) power. Street power is no different than what a typical home has for a source of electrical power. The power is raw (not conditioned) and has spikes and drops (low voltage). This is fine for things like lights, TVs and radios but definitely not for computers or networks.
A computer room will have UPS systems that condition the power and can maintain power on batteries for a short period of time (minutes) until emergency power kicks in.
Emergency power is usually a standby generator source, and it takes a few minutes to be brought up to peak requirements. LC is a maximum redundant electrical source. It uses multiple raw power inputs, including multiple-source electrical substations. The substations in turn are sourced by both continuous running generators and raw power from the area's electrical power grid.
Only a few years ago, health care networks were electrically powered by simple emergency power. When the conditioned grid power failed, then the UPS batteries held until emergency generators fired up. Although this sounds like a good redundant power plan, it's riddled with single points of failure. Now with more and more clinical devices requiring computers and networks, the health care network needs LC power.
Don't forget the core network
In the past, not many IT people paid a lot of attention to network core design. Most viewed the network core as the main pipe into which everything was plugged. The core became a single path of network cable that ran centrally through the institution. Well, that has changed dramatically. The core network now requires parallel and redundant devices.
The core should be fiber-based. Copper network cores are long gone. The minimum bandwidth of the core network should be 10 Gb. Interestingly, the core network doesn't need to be significant in length. It doesn't need to run the whole length of a building, from one end to the other. What is important is it provides redundant access and has dual paths and an electrical power source that's always available. Its location should be secure. The core network closets should not be closets and definitely not be the janitor's closets. In the past, network closets were never planned in the design of a building. Now there are specific standards for data communication rooms or telecommunications rooms (DCR/TRs). A DCR/TR should meet the minimum requirements of the ANSI/TIA/EIA-568-B standard.
There are many designs for parallel and redundant cores. A minimum design should be three DCRs with dual-core, redundant 10 GB switches. Each closet should be powered by nonassociated, separate LC emergency power. In the event of one room having any type of failure, the other two rooms can manage the network load. Performing maintenance on a health care network is equivalent to changing the oil on your car and rotating the tires while the car is speeding down the highway at 65 miles an hour. But this type of basic core design allows maintenance to be performed on the network equipment without taking out the entire network.
Al Gallant is the director of technical services at Dartmouth Hitchcock Medical Center in Lebanon, N.H. Let us know what you think about the story; email firstname.lastname@example.org.