For many companies the internet is no longer a nice extra. The accounting software runs on a cloud server, the phones use a SIP trunk and the branches connect to headquarters over VPN. With a single link, any provider outage or cable cut in the street stops the whole company. A second link is the cheapest insurance against this dependency, as long as it is designed correctly.
What Is the Difference Between Failover and Load Balancing?
- Failover: One link is primary, and the second link only comes into play when the primary goes down. This is simpler and more predictable.
- Load Balancing: Both links are used at the same time and connections are spread across them. If one goes down, all traffic moves to the other.
Load balancing usually works per connection, not per packet. A single file download does not get the combined speed of both links; instead, the many connections from your users are spread across the two links.
Choosing the Second Link
A second link is only a true backup if it does not fail at the same time as the first:
- Get it from a different provider.
- Ideally use a different physical medium, for example fiber or copper for one and a wireless or cellular connection for the other. Two links that come through the same cable and the same telecom exchange often go down together.
- Connect the modems and equipment for both links to a UPS.
Key Settings on the Firewall or Router
- Real link health monitoring: A lit modem light is not enough; the link may be up while the internet is not working. Configure the firewall to regularly ping several reliable destinations outside the provider’s network and take the link out of service if they stop responding.
- Policy-Based Routing: Send sensitive traffic such as phones, branch VPNs and cloud software over the most stable link, and low-priority traffic such as the guest network over the other link.
- Sticky Connections: Some systems, such as online banking portals and sites that are sensitive to IP addresses, log the user out if the source address changes mid-session. Keep each user’s or each destination’s connections on a single link.
- Separate NAT: Outbound traffic on each link must leave with that link’s own public address.
Do Not Forget Inbound Services
Failover is fairly simple for outbound traffic, but services that connect into the office from outside depend on the public IP address of one link:
- Branch and remote-worker VPNs: Define a backup tunnel on the second link’s address so that if the first link fails, the connection comes up over the other path.
- Services published from the office: If a service is reachable from outside by domain name, failing over requires changing its DNS record, and a short TTL makes that change take effect faster. Often, moving these services to a data center is the more reliable solution.
Test your failover. During a quiet period, unplug the primary link and check that internet, phones and VPN really work over the second link, and that everything returns to normal once the primary is reconnected.
Common Mistakes
- Two links from the same provider or the same physical path.
- Health monitoring based only on port status, not actual internet reachability.
- A backup link that has not been used for months, only to discover on the day of the outage that its subscription has expired.
- Users being logged out of banking systems because sticky connections are not configured.
Checklist
- Two links from two providers over two different media
- Health monitoring with pings to external destinations
- A routing policy for sensitive traffic
- Backup tunnels for VPNs
- Regular failover testing and monitoring of the second link’s status
ITPlus724 engineers can fit a second link into your network design and configure and test failover and load balancing on your firewall.