Most companies believe that their disaster recovery is handled. In my experience, the gap is usually not a backup. It is the site. The risk landscape is becoming more difficult. According to NOAA, the U.S. logged 23 separate billion-dollar weather and climate disasters in 2025, the third-highest annual total on record. Power and grid strain are rising alongside AI-driven demand, and the Uptime Institute’s 2025 survey found a sharp jump in operators who are very concerned about forecasting future capacity. Threats are increasing as capacity tightens.
A backup that lives in the same building, power grid, or floodplain as production is not disaster recovery. It is a second copy waiting to fail at the same moment as the first. Where you put your recovery infrastructure is the decision that determines whether the plan works when you need it. This article discusses how IT leaders should evaluate a disaster recovery location, what the real criteria are, and how to avoid the trade-offs that quietly undermine most DR strategies.
Why Site Selection Is the Foundation of Disaster Recovery
Business continuity depends on the assumption that when the primary site goes down, the recovery site stays up. This holds only if the two sites do not share the risks that take data centers offline. The same grid, metro, weather system, and fault line, and no redundancy has been built. You have created a single point of failure across two buildings.
The threats that justify a DR site are not all dramatic. Natural disasters make headlines, but they cause a small share of actual downtime. The everyday killers are hardware failures, human errors, ransomware, and regional power events. Outsourcing does not make the problem disappear: over nine years of tracking public outages, the Uptime Institute found that roughly two-thirds involved third-party providers, including cloud platforms, telecoms, and data center facilities. You can move where your systems run. You cannot outsource the responsibility of keeping them up. It starts with the physical and operational separation between production and recovery.
Where Will You Recover? Build, Provider, or Cloud
Before you pick a location, you pick a model. There are three. You build and run your own second data center. You rent recovery capacity from a provider through colocation. Or you recover into the public cloud. Each has a failure mode worth naming.
Roll your own. Total control, total cost. You buy, staff, and maintain a full second site that mostly sits idle. The hidden risk is the drift. Production changes, the DR site does not keep pace, and the gap surfaces at the worst possible moment.
Hyperscale cloud. Elastic and pay-as-you-go, which are attractive until you read the fine print. Recovery depends on a working network path to the cloud region, egress, and replication costs, and some on-prem workloads do not behave the same once they are running on someone else’s infrastructure.
Colocation with a provider. For most companies in these markets, this is the answer. You get dedicated space, power, and connectivity in a facility built for it, without the capital drag of a second on-premise data center or telco closet in an office building. The decisive advantage is predictability. You know exactly what capacity and performance you are getting. You also control the footprint, from a half cabinet for a lean DR posture to multiple cabinets and private cages as you grow. More on that below.
The Distance Question: How Far Is Far Enough?
There are no regulations that provide a specific number. U.S. financial regulators floated a requirement of 200 to 300 miles in the early 2000s but abandoned it as unworkable. Standards such as ISO 22301 and the NIST 800 series call for a “safe distance” without defining one. So the decision falls to you, and it comes down to physics and risk geography.
Too close and you share the disaster. A recovery site a few miles away protects you from single-building events, such as power loss or fire. It does nothing against a regional grid failure, a hurricane, a wildfire, or an earthquake that takes out the whole metro.
Too far and you lose your data. Distance creates latency, and latency dictates how replication can occur. The speed of light through fiber adds roughly 1 millisecond of round-trip latency for every 100 miles before accounting for switches, firewalls, and the routing reality that fiber never runs in a straight line.
A practical floor is approximately 30–100 miles for separation from most localized and many regional events. The ceiling depends entirely on the replication model, which is the part most teams underweight.
Latency and Replication: The Tradeoff That Defines Your RPO
Synchronous replication confirms every write operation at the recovery site before the transaction is completed at production. The payoff is a Recovery Point Objective of zero. No data loss occurs during failover. The cost is that every write waits on the round trip, so the sites must be close. Most workloads cannot tolerate more than approximately 5 ms of round-trip latency, which practically caps synchronous replication at metro-to-regional distances, often cited as approximately 150 to 200 miles depending on the application.
Asynchronous replication confirms the writing locally and ships them to the recovery site in the background. It tolerates near-unlimited distances, which makes it the standard for long-haul DR. The trade-off is a replication lag window. If production fails, anything that has not yet been copied is lost, and that lag defines your RPO.
The right answer is workload-specific. A transactional database with a zero data loss requirement needs synchronous replication and a site inside the latency window. A file server with a 30-minute RPO is acceptable for async across the country. Mature shops often run both synchronous to a nearby site for zero RPO and asynchronous from there to a distant third site for regional protection. Distance and replication are not separate decisions. They are the same decisions.
A real number makes this concrete: I have tested 4–5 milliseconds of round-trip latency between St. George, Utah and a major Las Vegas carrier hotel over diverse fiber. That sits well inside the synchronous window, which is the whole reason a site a couple hours away can protect a Vegas production environment without forcing you onto async and accepting data loss.
The Seven Criteria: What Actually Makes a Good DR Site
Distance and latency are the main issues. However, they are not the whole list. Seven criteria separate a real DR site from a second copy waiting to fail:
Separate power grid. Different utility, different substation, ideally different generation. A recovery site on production’s grid fails the moment that grid does.
Different disaster profiles. Low seismic activity, no hurricane exposure, and minimal flood and wildfire risks. The best DR geography is the type that rarely makes the news.
Different state. Cross a state line and you often change the regulatory and tax picture, the utility, and the disaster profile all at once. Putting real distance and a separate jurisdiction between production and recovery is what keeps one regional event from taking down both.
Carrier-rich connectivity. Multiple fiber paths and diverse carriers protect the replication link itself. Cloud on-ramps are important for hybrid IT environments.
Resilient infrastructure. Redundant power, cooling, and networks, so that the site can be maintained without going dark. A DR site that requires a maintenance window is a DR site with gaps.
Accessibility and remote hands. If a disaster forces a physical response, can your team actually reach the site? Highway access, a nearby airport, hotels, and basic amenities are all important when someone must be on-site for hours or days. So does safety. You want a location in a good area, not somewhere your staff will not want to be at 2 a.m., and when no one can travel, on-site remote hands are the difference between a clean failover and an extended outage.
Proven failover, not assumed failover. A site that has never been tested is a guess. Most companies do not run real failover drills, which is how recovery plans fail when they are needed. The right partner makes testing routine, with managed support to run a true drill instead of a paper one.
How Many DR Sites, and What Footprint Do You Need?
Most companies require a well-chosen recovery site. According to Forrester and the Disaster Recovery Journal, 93 percent of companies with DR have at least one site, and 34 percent run more than one. A second site earns its cost when a zero-RPO requirement forces a near-synchronous pair plus a distant async target or when regulation and customer commitments demand geographic spread.
The footprint scales with what you are protecting. A small business replicating core systems might require a half cabinet. A mid-market enterprise running critical applications and storage usually lands at a full cabinet or a few cabinets. Larger or high-density environments with strict security standards usually land in cages and private suites. Size for what has to come back online to keep the business running, not a rack-for-rack mirror of the production. Moreover, most backup jobs run overnight, so the site and its links have to handle that nightly surge, not just the daytime average.
Active-Active vs. Active-Passive
Both terms describe how you run a workload across two sites. The difference is whether the second site does work or just waits.
In active-passive mode, production runs the workload, and the recovery site waits on standby, ready to take over if production fails. In active-active, both sites run live and share the load. If one fails, the other absorbs everything with little or no failover delay.
Active-active provides the strongest availability and the lowest RTO, but it requires low latency between sites, tight synchronization, and more infrastructure. It’s for workloads that genuinely can’t go down.
For most mid-market environments, a well-designed active-passive setup with the appropriate replication model achieves the right balance of cost and resilience.
Why Location Matters More Than Ever: The Capacity Crunch
Data center capacity is limited across nearly every major market. AI workloads are pulling power and space faster than new facilities come online, and the obvious DR markets, the large metros, are exactly the ones running out of room and carrying the highest disaster risk.
When a regional disaster occurs, every client of a generic DR provider reaches for the same pool of capacity simultaneously. If you oversubscribe that pool, your recovery will degrade at the moment you need it the most. The fix is a dedicated space in a market with room to scale, secured before the crisis, not during it. The window is closing. A 2026 U.S. Chamber of Commerce Foundation survey found that 69 percent of small businesses still have no disaster plan at all, which means a rush for capacity is exactly what the next big event will trigger.
For the West Coast and Southwest, the Mountain West is a clear choice. For companies in Las Vegas, Phoenix, Reno, Los Angeles, and Salt Lake City, a recovery site in southern Utah checks the boxes that matter: low natural disaster risk, a stable grid, competitive power rates, carrier-rich connectivity, and low latency back to those markets. Far enough to sit outside the disaster radius of any single metro. It is close enough to keep replication fast. In the last year alone, I have helped move two large customers out of Las Vegas data centers, and the pattern driving those moves, cost and capacity pressure in a high-risk metro, is exactly what is pushing DR planning toward markets like this one.
Best Southwest and West Coast Disaster Recovery Options
ValorC3’s St. George, Utah data center was purpose-built for this role. The facility is roughly two hours or less from the Las Vegas Strip by road, which keeps it inside a workable latency window for the Vegas market while placing it on a separate grid, in a separate state, with a completely different disaster profile. Southern Utah is low-seismic, has no hurricane exposure, and runs on stable and competitively priced power.
The site backs the geography with real infrastructure: a dedicated municipal utility substation, N+1 redundant generators, redundant UPS and cooling, dual-interlock clean-agent fire suppression, and a carrier-neutral environment with multiple fiber paths and cloud on-ramps. The colocation footprint scales with your DR posture. Start with a half cabinet for a lean recovery site, grow into multiple cabinets as you protect more of the environment, and move into a private cage or suite when you need a dedicated, secured space of your own.
ValorC3 also runs interconnected data centers between St. George and Boise for true multi-site separation, and Valor Cloud is available in both locations, with managed DRaaS and BaaS built on Veeam. Colocate your own gear, replicate to a second ValorC3 site, or hand off backup and recovery as a managed service, all under one partner.
Capacity in the right markets is constrained and is being filled. If you are rethinking your disaster recovery strategy or building one for the first time, the time to secure space is before you need it.
Disaster Recovery Site Selection: Common Questions
What is the ideal distance for a disaster-recovery site?
There is no single number. A practical floor is 30–100 miles, far enough to escape a localized event. The ceiling depends on the replication model. Synchronous replication, which results in zero data loss, generally requires sites within approximately 150–200 miles. Asynchronous replication tolerates greater distances but accepts a small data-loss window.
How many disaster recovery sites does a company require?
Most companies require a well-chosen recovery site. A second site is justified when there is a zero-RPO requirement that calls for a near-synchronous pair plus a distant asynchronous target, or when regulations and customer commitments require a geographic spread.
What is the typical footprint of a DR site?
It scales with what you are protecting. A small business may require only half a cabinet. A mid-market enterprise typically lands at a full cabinet or a few cabinets. Larger or high-density environments move into cages and private suites. Size for what has to come back online to run the business, not a full mirror of production.
What is the difference between active-active and active-passive?
In active-passive mode, production runs the workload, and the recovery site waits on standby. In active-active, both sites run live and share the load; thus, if one fails, the other absorbs everything with little or no failover delay. Active-active delivers the lowest recovery time but demands low latency, tight synchronization, and more infrastructure than the other strategies.
What is the difference between RTO and RPO?
The recovery point objective (RPO) is the amount of data that can be lost, measured as a time window. The recovery time objective (RTO) is the time that a system can be offline before it is restored. Your replication model and site choice are what make these targets achievable.
Brooks Rashid, Enterprise Account Executive
With more than 15 years in enterprise infrastructure spanning colocation, cloud, and managed services, Brooks helps companies across the Southwest and West Coast design disaster recovery and colocation strategies that perform under pressure. He has guided organizations through site selection, migration, and the decisions that keep critical systems online. He is based in the Salt Lake City area.