In many ways, site selection for a health care organization is completely the opposite. In a health care environment, the facility and labor costs are still important but are generally secondary to providing critical care. Common across the health care industry is the need to provide critical care for patients and clients. This is an obvious requirement for hospitals, but it is also necessary for clinics and health care organizations providing services like claims approvals, nurse-by-phone consultations, suicide hotline consultation, etc. Highly available infrastructure is needed for the health care offices because of the criticality of the services provided and the expensive labor force provided by the doctors and nurses. Generators, UPSs, redundant power, and redundant network connections provide the backbone for this high availability.
The presence of the highly available infrastructure for an office space and the proximity of administrative staff makes adding a data center in the same location at build-time an obvious decision. There are clear synergies by colocating a data center in a highly available office space because construction costs are minimized. However, retrofitting a highly available office space for a data center move is also an easily justifiable option since infrastructure expenses are very low because of the baseline infrastructure that already exists. If a highly available office exists, there is very little financial reasoning to put the data center elsewhere.
Balancing these factors when selecting data center locations is a regular activity for IT veteran Joseph Steele. Mr. Steele is currently vice president of Infrastructure at MedSolutions and was previously vice president of IT at APS Healthcare. The criteria that are most important to Mr. Steele when selecting a site selection all revolve around mediating risks. Ensuring high availability for critical services is an absolute requirement. Below is a list of his selection criteria for choosing a data center location.
Select locations that are in different geographic regions to avoid outages from a regional disaster. A safe rule of thumb is to keep data centers at least 500 miles apart. A secondary or tertiary data center in a separate region can provide business continuity during a major disaster like Hurricane Sandy or Hurricane Katrina.
Connectivity and Latency
Data centers need to be far enough apart to avoid regional disasters while being close enough to minimize latency. Latencies of well below 100 ms are needed to ensure customer-facing applications operate without lag. Low latencies are also needed to ensure effective asynchronous data replication for storage area networks and disk-based backups. For instance, a primary data center in Chicago and a secondary in Dallas are roughly 900 miles apart, but they still maintain a latency of only 20 ms if they’re both on the same AT&T backbone.
Allocating multiple Tier 1 and Tier 2 carriers provides high network availability while ensuring there is a backup circuit in case of failure. However, having multiple network carriers drives up the cost, especially if there is a large distance between data centers.
Some locations have their own unique issues. For instance, maintaining redundant carriers is particularly important in Puerto Rico because network outages are common. Outsourcing data center services to India comes with latency problems that can often exceed 150 ms. Latency this high can prevent even highly resilient Citrix remote desktop connections and Citrix published applications from working.
Common across the health care industry is the need to provide critical care for patients
Natural disasters like lightning and tornadoes are common, but strikes in any individual area are rare enough to be essentially ignored from a risk perspective. Hurricanes are earthquakes are more problematic, although flooding is by far the biggest problem. These risks can be avoided by selecting sites that are 100 miles from the coast, that are in seismic zones 0 or 1, and that are outside of the 100-yr flood zone. Even knowing the risks, data centers continue to be placed in high-risk areas. Consider Silicon Valley, which is less than ideal for data center site selection. The area is 30 miles from the coast, is in an earthquake zone 4 (the highest risk), is about 50% in the 100-yr flood plain, and has real estate and electricity that is among the most expensive in the country.
Proximity and High Availability
Colocating the data center in a building that has high availability requirements reduces the investment needed to support the data center. Additionally, the response times for resolving data center related problems are shortened if staff from the facilities, network, and systems engineering teams are located in the same place.
The idea of selecting a cold site for pure disaster recovery is now an obsolete concept. The onset of virtualization, converged networks, migration to SIP and VoIP communications, and the growth of private and public clouds has simplified the deployment of live production systems across both the primary and backup data centers. Instead of letting the backup site lay fallow until a disaster, site utilization can be maximized by running load-balanced production systems across data centers and across the cloud to ensure high availability. A side benefit of load balancing applications across sites is that it also increases regional application performance by putting the applications closer to the customers.
The portability of virtual systems dramatically reduces the effort needed to migrate between data centers. Colocation space at hosting providers can be considered as almost a pure commodity because of the ease of moving systems in and out of the space. Site selection for buy/build/lease decisions are much easier when the systems that need to be supported are highly portable.