Blog
Super Intelligent DNS with CloudFloorDNS
October 5, 2026
Ordinary DNS is a phone book. Ask it for www.yourcompany.com and every visitor gets the same address, every time, whether that server is healthy, overloaded, or on the other side of the planet. Super Intelligent DNS is our name for DNS that checks before it answers: it knows which of your servers are up, where each visitor is coming from, and which server will serve them best.
“Intelligent” gets attached to a lot of products right now, so this post stays concrete: how the decision is made, what happens when a server dies at 2 a.m., how we compare with Route 53, Cloudflare, NS1 and Azure, and the trade-offs a sales page usually leaves out. All of it runs on our network today.
What makes DNS intelligent?
A plain DNS record answers one question: what is the address? Super Intelligent DNS asks four more first.
Netmon tests each server from up to seven locations, using the protocol your users use: HTTPS for a website, SIP for a phone system, SMTP for mail.
Geo DNS Director maps each query to a country or a region you define, using the visitor’s own subnet when the resolver shares it.
Load balancing spreads visitors across a pool of servers by rotation, by weight, by geography or by measured latency.
When three or more monitoring nodes agree a server has failed, DNS Failover moves its traffic to your backup in as little as 60 seconds.
None of these ideas is new. The intelligence is in combining them, so each answer reflects the live state of your infrastructure rather than whatever someone typed into a zone file last spring. The industry calls this global server load balancing (GSLB) or DNS traffic steering.
What happens inside a single lookup
Here’s the path one query takes, from the moment a visitor’s resolver asks for your hostname.
The query reaches the nearest node. Our dual-stack, DDoS-protected anycast network has 15+ points of presence on four continents.
We work out where the visitor is. We use the visitor’s subnet if the resolver sent an EDNS Client Subnet hint, or the resolver’s own address if not, and match it to a country or one of your region groups.
We pick that region’s pool. For example, Europe gets your London and Frankfurt servers, North America gets Virginia and Dallas, and everyone else gets your default pool.
Failed servers are already out. Netmon has been testing every pool member all along, and any server that three or more monitoring nodes agree is down has already been switched out. The answer never waits on a live test.
We choose among the healthy servers by round-robin, by weight or by latency, depending on how you’ve set up the pool.
The answer goes out with a short TTL, so resolvers check back soon and the next change, a failure or a recovery, reaches visitors quickly.
Nobody has to be awake for any of it. That’s the point.
What it looks like in practice
Your main site goes down overnight
Netmon checks your web server every minute. When three or more monitoring nodes confirm the failure, the record switches to your standby server or a maintenance page, and Netmon alerts you by email or SMS. Nobody has to log in to make it happen.
Overseas customers say the site is slow
Put a server in Europe or Asia and let Geo DNS Director send visitors there by country while everyone else stays on your main site. Geo pools support failover too, so if the overseas server goes down, its visitors can be sent to a working server instead of an error.
You’re rolling out a new server
Add it to the pool with a small weight so it takes, say, one visitor in ten. Watch your logs, then raise the weight. If it starts failing health checks, it drops out of the pool on its own.
Your phone system has to stay up
Geo DNS Director can serve location-specific SRV records, so softphones and SIP trunks register with the nearest gateway, and Netmon’s SIP checks can fail a dead gateway over to a working one. Our VoIP failover page has the details.
Under the hood
For the engineers, here’s each building block and the numbers behind it.
| Building block | What it does | Details |
|---|---|---|
| Anycast DNS network | Answers from the node closest to the visitor | 15+ PoPs on four continents, IPv4 and IPv6 at every node, built-in DDoS protection, 100% uptime SLA |
| Geo DNS Director | Chooses answers by visitor location | Country or custom region groups, EDNS Client Subnet, A, CNAME and SRV records, plus geo URL and path forwarding |
| Netmon monitoring | Tests every server, continuously | 20+ test types including HTTP(S), TCP and UDP connect, ping, SSH, SIP, SMTP, IMAP, POP3, DNS and SSL expiry; every 10 seconds to every hour, from up to 7 monitoring locations, depending on plan |
| DNS Failover | Swaps failed servers out of the answer | A, CNAME and ALIAS records; acts when 3 or more monitoring nodes agree; in as little as 60 seconds |
| DNS Load Balancing | Spreads traffic across a pool | Round-robin (free on every plan), weighted, geographic and latency-based; Netmon adds and removes pool members |
Why we wait for a quorum
One monitoring node can be fooled by a routing problem between it and your server, and failing over on a false alarm is an outage of its own. Requiring three or more nodes to agree stops one bad network path from moving your traffic. The big platforms make the same call in their own way: Route 53 marks an endpoint unhealthy when 18% or fewer of its checkers say it’s healthy, and Cloudflare goes by a majority of regions.
Nearest isn’t always fastest
A server 500 miles away on a congested link can lose to one 2,000 miles away on a clean one. That’s why our pools can balance on measured server latency, not just geography, and why Geo DNS Director can fail a pool over on latency as well as on downtime.
Location accuracy depends on the resolver
We see the visitor’s resolver, not the visitor, and ISP resolvers usually sit close to their customers. Google Public DNS also passes a shortened copy of the visitor’s subnet (EDNS Client Subnet) to nameservers that support it, as ours do. Cloudflare’s 1.1.1.1 deliberately doesn’t, for privacy, so its users are placed by the 1.1.1.1 node that asked, which is normally nearby. That’s true for every GeoDNS provider, not just us, and it’s why country and region rules work better than trying to be city-precise.
Setting it up
Create your account and add your zone
Sign up at panel.cloudfloordns.com, or send us your zones and we’ll import them. Steered hostnames must be answered by our nameservers, so point your domain at CloudFloorDNS at your registrar once your records are in place.
Add a Netmon check for every server
Match the test to the service: HTTPS against a real page for a website (a ping only proves the machine is on), SIP for phones, SMTP or IMAP for mail. One minute is a sensible interval.
Build your pools
On each hostname you want to steer, create a failover pair or a load-balanced pool and attach each member’s Netmon check. Give bigger servers bigger weights.
Add geography
In Geo DNS Director, map countries or groups of countries to pools, and always set a default pool for everyone else.
Set TTLs on purpose
Use 60 to 300 seconds on steered records, and leave records that never move at their usual longer values.
Test it before you need it
Query our nameservers directly and simulate visitors from other countries. Then stop the web service on a test server (not the whole machine) and watch the answer change.
# What are we answering right now?
dig +short www.example.com @<your-cloudfloordns-nameserver>
# Simulate a visitor in the UK (EDNS Client Subnet)
dig +short www.example.com @<your-cloudfloordns-nameserver> +subnet=81.2.69.0/24
# Confirm the TTL resolvers will cache
dig www.example.com @<your-cloudfloordns-nameserver> +noall +answerWhat it costs
There’s no separate price tag. The pieces are built into our Managed DNS plans:
- Round-robin load balancing is free on every plan.
- DNS Failover is an option on the $8/month Developer plan and included from Bronze ($50/month, two one-minute failovers) up to Platinum ($300/month, eight).
- GeoDNS is an option on Developer and Bronze, with one GeoDNS host included on Silver ($80/month), two on Gold ($120/month) and four on Platinum.
- Standalone monitoring for other servers and services is free for one hourly check, with paid Netmon plans from $5/month. The $60/month Enterprise monitoring plan checks every 10 seconds from all seven locations.
- Larger portfolios get custom yearly pricing. Ask us for a quote.
How it compares
Every major cloud platform sells some form of DNS traffic steering. Here’s how the best-known options line up for a typical setup: a few hostnames, two or three servers each, with health checks and geographic routing.
| Provider | Steering methods | Health checks | Pricing (list) | Worth knowing |
|---|---|---|---|---|
| CloudFloorDNS | Geo (country or custom regions, with EDNS Client Subnet), failover, round-robin, weighted, latency; geo SRV and URL forwarding | 20+ test types; every 10 s to 1 h, depending on plan; 3 or more monitoring nodes must agree before failover | Flat monthly plans: failover included from $50/mo, GeoDNS hosts from $80/mo, round-robin free | Pure DNS, so no proxy in your traffic path; works with any host, cloud or registrar |
| AWS Route 53 | Failover, geolocation, geoproximity, latency (by AWS Region), weighted, IP-based, multivalue; Traffic Flow for chained policies | Every 10 s or 30 s; unhealthy when 18% or fewer of checkers report healthy | Pay per use: $0.50 per zone/mo, $0.70 per million geo queries, $0.75/mo per non-AWS health check plus $2/mo per optional feature, $50/mo per Traffic Flow policy record | Very capable, but costs grow with queries, checks and features. Simplest when your servers already live in AWS |
| Cloudflare Load Balancing | Failover order, weighted random, geo, dynamic (latency), proximity, least outstanding requests | HTTP(S), TCP, ICMP, UDP-ICMP, SMTP; 60 s by default, faster checks depend on plan; all-region checks are Enterprise-only | Add-on starting at $5/mo, rising with the options you add | Load balancers live inside a Cloudflare zone; keeping your DNS elsewhere needs a CNAME setup (Business or Enterprise) or a secondary setup |
| IBM NS1 Connect | Filter chains (geo, ASN, health, weighted and more); Pulsar real-user-measurement steering | Monitors included per plan: 2 on Essentials, 2 to 100 on Standard | Essentials from $99/mo (1 filter chain), Standard from $349/mo, Premium by quote; Pulsar is a paid add-on | Highly programmable and aimed at large enterprises; entry plans cap filter chains and monitors |
| Azure Traffic Manager | Priority, weighted, performance, geographic, multivalue, subnet; nested profiles to combine methods | HTTP, HTTPS, TCP; every 30 s, or 10 s with fast probing; HTTPS probes don’t validate the certificate | Pay per use: $0.54 per million queries, $0.54/mo per external endpoint check, plus $2/mo each for fast or HTTPS probing | Reached by CNAME to a trafficmanager.net name, so using it on an apex domain means hosting that domain in Azure DNS |
The short version: AWS and Azure make sense when everything already runs there, NS1 when you need deep programmability and have an enterprise budget, and Cloudflare when you already proxy your sites through it. If you want steering at the DNS layer that works with any host, cloud or registrar, at a flat monthly price, with people who pick up the phone, that’s what CloudFloorDNS is built for.
Things worth knowing before you turn it on
DNS steering is only as fast as caching allows. We can switch a record in about a minute, but resolvers that cached the old answer keep it until its TTL runs out, and some applications hold on to answers even longer. Keep steered TTLs short and expect a small tail of stragglers after any change.
- Test what users actually use. A server can answer ping while its application throws errors. Point HTTPS checks at a page that touches your database.
- Make sure the backup can carry the load. Failing over to half the capacity, or to yesterdays data, swaps one problem for another.
- Plan the trip back. Know what has to be true before the primary takes traffic again, and test the return as well as the failover.
- Zone transfers copy records, not decisions. If you also run a secondary DNS provider (here’s why you might), it serves your zone’s static records, not our geo and failover logic, so set those static values to a sensible default.
Frequently asked questions
Is Super Intelligent DNS a separate product?
No. It’s Geo DNS Director, DNS Failover, Load Balancing and Netmon monitoring working together on our anycast network. You switch on the pieces you need, on the hostnames that need them.
Do I have to move my DNS to CloudFloorDNS?
The hostnames you steer have to be answered by our nameservers. Your domain can stay at its current registrar, and we’ll import your existing zones for you.
How fast is failover, really?
As little as 60 seconds after three or more monitoring nodes confirm a failure, plus the time resolvers take to drop their cached copy. With a 60-second TTL, most visitors reach the backup within a few minutes.
Is this AI?
No. Every decision follows rules you set and can read: health checks, location, weights and measured latency. When traffic moves, you can see exactly why.
Make your DNS think before it answers
Start with a free account and put a Netmon check on your most important server, or send us your domain list and we’ll import your zones and help you build your first geo and failover pools.
Prefer to talk it through? Call +1 781 373 5823 or email sales@cloudfloordns.com. Existing customers can reach support@cloudfloordns.com.
Related reading: Why CloudFloorDNS for Managed DNS, Website & Infrastructure Monitoring Guide and Why Should I Use Secondary DNS?