The questions people actually ask before choosing a facility, answered directly — including the ones where the answer is no.
Colocation means placing your own servers, storage and network equipment in someone else's purpose-built data center. You own the hardware; the facility provides the building, conditioned and backed-up power, cooling, physical security, fire suppression and carrier access.
It sits between running equipment in your own office and renting virtual capacity from a cloud provider. It suits workloads that are steady, that you want to control directly, or that need to live in a specific physical place.
A few rack units inside a shared locking cabinet. There is no minimum that quietly excludes small tenants — a single server is a reasonable order. From there the standard steps are partial cabinet, half cabinet at around 20U, full 42U cabinet, and private cage for multi-cabinet deployments. See the rack space page for what each looks like here.
Price depends on three things: how much space you take, how much power you commit to, and how long your term is. Power is usually the larger driver — two cabinets drawing 2 kW each cost less to serve than one cabinet drawing 8 kW.
Bandwidth is typically priced separately, and remote hands is quoted per incident or as a bundled allowance. We quote against your actual numbers rather than a rate card, and every allocation and overage rate goes in writing before you sign anything.
Space in a locking APC Netshelter SX 42U cabinet, a committed power allocation backed by UPS and diesel generator, N+1 cooling, environmental monitoring, clean-agent fire suppression, 24/7 access under three-layer key fob control with logging, continuous video surveillance, and network handoff to your choice of the five in-building carriers. Bandwidth, IP allocations above your baseline, cross-connects and remote hands are quoted separately.
For a standard cabinet using carriers already present in the building, days rather than months. The sequence is agreement, cabinet and circuit allocation, shipping or hand-carrying equipment, racking, network handoff and access fob issue. If you need a new carrier circuit provisioned, that is the long pole and the timeline belongs to the carrier rather than to us.
Yes. Short-term and swing space is available and we would rather fill it than hold it empty. Common situations include migrating between facilities and needing parallel-run capacity, bridging an unacceptable renewal at another provider, rebuilding after a security incident, or standing up infrastructure for a divested business unit on a transaction timeline.
This facility carries 40 kW of UPS-protected capacity today, expandable to 80 kW on the installed APC Symmetra PX frame. At conventional density of 2 to 3.5 kW per rack that supports roughly twelve cabinets now and roughly twenty-four after expansion. Individual cabinets at 8 to 10 kW are possible for a small number of racks.
If you need more than about 10 kW in a single cabinet, raise it in your first message so we can tell you honestly whether it fits, fits after an upgrade, or does not fit here at all.
Five: Comcast/Xfinity, WOW, LCUB, Iris Networks and AT&T. Additional providers operate in close proximity. You can take bandwidth through us or bring your own carrier and we provide the handoff.
BGP is supported, including announcing your own address space if you have an ASN, and an autonomous IPv4 C-class block is available for assignment. Carrier-provided DDoS mitigation with traffic black-holing and automatic BGP reroute is in place.
Tier II means redundant capacity components — N+1 power and cooling, so a single component failure does not take the floor down. Tier III adds concurrent maintainability: any component can be taken out of service for maintenance without interrupting the load, which requires redundant distribution paths as well as redundant components.
This facility is Tier II with N+1 redundancy, not 2N and not concurrently maintainable, so maintenance windows exist. For most colocation and for cold or warm disaster recovery tiers that is immaterial. If your workload cannot tolerate a scheduled window, you want a Tier III facility, and there are good ones in this market.
N is the capacity needed to carry the load. N+1 means one additional unit beyond that, so if one fails the remaining units still carry everything. This facility runs N+1 on both the UPS and the Mitsubishi cooling systems. It is different from 2N, which duplicates the entire system end to end and belongs to higher tier classifications.
On-site staff performing physical tasks on your equipment when you cannot be there — power cycling a device, swapping a cable or a drive, changing media, reading a console, or putting eyes on a failed component. For tenants located hours away, and for disaster recovery footprints in particular, remote hands is usually what makes the arrangement practical.
Yes. Nominated staff hold Kantech key fobs giving round-the-clock access to your own space, with all entries logged. Access lists are maintained in advance so nobody has to arrange permission during an incident.
Distributed inference at conventional density, yes — that is a genuine fit, and inference is now the majority of ongoing AI compute load precisely because it has to sit near users.
AI training, no. Modern training racks draw more power individually than this entire floor carries, which is why the industry is moving to 800 VDC distribution architectures to feed them. A facility this size should be honest about which half of that market it serves.
Industry usage puts an edge data center at roughly one rack up to around fifty, serving workloads that need to sit physically close to users. That describes this facility accurately.
It is worth distinguishing from a micro data center, which in dominant industry usage means a self-contained one-or-two-rack enclosure sold as a hardware product by vendors such as Schneider, Vertiv and Rittal. That is a different thing entirely, and not what a 1,209 sq ft purpose-built building is.
Far enough that a single event cannot affect both sites, which in practice means a different power grid, different flood and storm exposure, and network paths that do not share the same conduits. Many organisations use a rule of thumb of at least 100 to 250 miles, though the real test is hazard separation rather than a specific number.
Working against that is latency and travel time: the further apart, the harder synchronous replication becomes and the longer it takes staff to reach the site. Knoxville sits roughly 150 miles from Atlanta and Nashville, around 200 from Charlotte, and 500 or more from the Florida coast — inside a workable band for most Southeast primary sites.
A cold site holds equipment powered down in secure, climate-controlled space; recovery takes days. A warm site keeps equipment powered, connected and receiving scheduled replication, so declaring means activating rather than rebuilding; recovery takes hours. A hot site runs production-capable equipment continuously with near-continuous replication, recovering in minutes.
Cost rises steeply across the three, and most organisations find on inspection that different systems in their estate belong in different tiers. All three are available here — see the disaster recovery page.
RTO, recovery time objective, is how long you can be down before the impact becomes unacceptable. RPO, recovery point objective, is how much data you can afford to lose, measured backwards from the moment of failure.
Together they determine which standby tier you need and how frequently you replicate. Setting them honestly is the single most useful thing to do before shopping for a DR site, because it converts a vague requirement into a specification.
Keep three copies of your data, on two different types of media, with one copy off-site. Colocation space is a common way to satisfy the off-site element with infrastructure you control, rather than relying entirely on a cloud backup provider. In ransomware scenarios the off-site copy matters most when it is genuinely unreachable from the production network.
Yes, and it is now one of the more common reasons tenants take standby space. Backups reachable from your production network are reachable by an attacker as well.
A separate building on a separate network, with independent access control and no trust relationship with your production domain, gives you somewhere to hold immutable copies and somewhere clean to rebuild into while the primary environment is quarantined or under forensic examination. Cold and warm tiers both serve this well — the separation is what matters, not the readiness level.
Safer than the Gulf Coast, and we set out the full picture rather than the marketing version. There is no storm surge or coastal wind exposure here, and tornado frequency is lower than the southern Plains or the Dixie Alley core.
But tropical remnants do reach East Tennessee — Helene caused federally declared disasters in nine counties here in 2024, though Knox County was not among them. And Knoxville sits within the Eastern Tennessee Seismic Zone, the second-most active seismic zone east of the Rockies, which produces frequent small events though nothing above roughly magnitude 4.8 in the instrumental record. The disaster recovery page covers both in detail.
No. This facility holds no compliance certifications and is not pursuing them. Any listing you find elsewhere claiming otherwise is inaccurate.
What we do provide is detailed documentation of the physical and environmental controls in place — three-layer access control with logging, continuous video surveillance, clean-agent fire suppression, locking cabinets and cage partitions, environmental monitoring. Many organisations in regulated sectors can use documented physical controls within their own compliance programme, but whether that works for your specific obligations is a question for your auditor, and it is better asked now than later.
For steady, predictable workloads that do not need to scale up and down, it frequently is — particularly once data transfer charges and licensing are counted. For spiky or unpredictable workloads, the cloud's elasticity is genuinely worth paying for.
The honest comparison includes hardware capital or lease cost, colocation fees, bandwidth, and your own staff time, set against cloud compute, storage, egress and support. Workloads with large or constant outbound data transfer tend to show the sharpest difference.
What power is committed versus metered, and what is the overage rate. What is the actual kW ceiling per cabinet. Which carriers are physically in the building, and what does a cross-connect cost. What is the tier classification, and are maintenance windows required. Who do I call at 3am, and is it a person or a queue. What is included in remote hands and what is billed. Is there room to grow, and on what terms. What compliance certifications do you actually hold, if any.
Any provider that answers all of these directly is worth talking to further. Any provider that answers none of them on their website is telling you something.
Ask it directly. We would rather tell you this facility is the wrong fit than waste a month of your evaluation.
Ask a question