Backup Internet for the Office: A Failover Setup Guide for SMBs

In a modern office, "the internet is down" no longer means a minor inconvenience. E-invoicing stops, cloud accounting and CRM screens will not load, IP phones fall silent and card terminals cannot authorise payments. Staff are at their desks, but the business is effectively closed. Despite this, most small and mid-sized companies still run on a single line and file each outage under bad luck — when a second line and a properly configured automatic failover turn the same risk into a modest, predictable monthly cost.

Failover vs Load Balancing

Failover is an active-passive arrangement: when the primary line fails, traffic moves automatically to the standby line, and moves back (failback) once the primary recovers. Load balancing is active-active: both lines carry traffic all the time. Balancing adds capacity, but configured carelessly it breaks sessions on banking sites and some corporate applications. For a business whose goal is outage protection, the starting point is always the simple, predictable failover; balancing is a later step, adopted only when measured demand justifies it.

The Four Components

1. A Dual-WAN Firewall or Router

The intelligence that performs the switch is a device that can manage two WAN connections simultaneously. Practically every business-class firewall and mid-range-plus router supports dual WAN; consumer modems do not. Whether your current device has this capability is the first question of the project.

2. A Second Line Independent of the First

The backup line must be chosen so that it does not fail together with the primary — the diversity principle below governs exactly this.

3. Health Checks

The device must also catch the failure mode where the primary link looks "up" but carries no data. Switchover decisions are therefore tied not to the physical link state but to periodic test requests (ping or DNS lookups) toward the outside world. A check that only pings the modem cannot see an outage on the ISP's side — defining multiple external targets is the sound practice.

4. A Failback Rule

When the primary recovers, the timing of the return is a deliberate decision. An instant return produces ping-pong switching on a flapping line; healthy setups wait until the line has been stable for several minutes before moving traffic back.

Choosing the Second Line: the Diversity Principle

• Provider diversity: buy the two lines from different ISPs, so one provider's regional outage cannot take both down.
• Infrastructure diversity: two lines with different ISP logos may still ride the same physical plant. The question to ask the provider is blunt: "Do these two circuits leave the same exchange and the same street cabinet?"
• Technology diversity: the strongest combination pairs a fixed line with an entirely separate carrier technology such as 4G/5G — an excavator cutting the fibre does not touch the cell tower.

What the Switchover Feels Like

Failover completes within seconds, but "nobody notices anything" is the wrong expectation. Services behave differently during the transition:

Service	Behaviour at switchover	What users experience
Web and cloud apps	New connections use the backup line	A brief slowdown, a page refresh
Email	Clients retry automatically	Usually unnoticed
IP telephony / VoIP	Active calls drop; new calls go out on the backup	A call cut short, redialled
Site-to-site VPN tunnels	Tunnel drops and re-establishes on the new path	A one-to-two-minute gap
Remote desktop sessions	Session drops and reconnects	Unsaved screen state lost

Explaining this table to the team in advance prevents the "the backup line doesn't work" perception on outage day: it works — the transition simply makes a few seconds of waves.

The Static IP and DNS Trap

If anything reaches your office from outside — a camera recorder, a locally hosted application, the head-office end of a branch VPN — that access is usually pinned to the primary line's static IP. When traffic moves to the backup, the outside world still looks for you at the old address and fails. The remedy depends on the scenario: static IPs on both lines with critical services defined on each, dynamic DNS, or moving externally exposed services to the cloud side. If the "what connects to us from outside?" list is not drawn up before the installation, failover will look flawless from inside while every external party experiences an outage.

Installation Steps

• 1. Inventory and risk list: write down the internet-dependent processes (payments, telephony, cloud apps) and the systems reached from outside. A rough cost-per-hour-of-outage figure sharpens the investment decision.
• 2. Procuring the second line: choose by the diversity principle; delivery typically takes one to three weeks.
• 3. Device configuration: define both WANs, set health-check targets, tune the failback delay.
• 4. Controlled test: outside working hours, physically pull the primary line; observe and record switchover time, service behaviour and the return.
• 5. Monitoring and alerts: make sure the switchover event lands in administrators' inboxes — otherwise the office can run on the backup line for days with nobody aware, and the primary's fault never gets reported.

The Monthly Drill

Failover is an assurance on the day it is tested, not the day it is installed. Once a month, run a five-minute controlled drill: shut the primary line, measure the switchover, exercise the critical applications on the backup, and log the result in a single line. The same discipline that applies to UPS and generator testing on the power side applies to line redundancy.

How Yamanlar Bilişim Approaches These Projects

We start failover projects with discovery rather than a device catalogue: your internet-dependent processes, your current device's dual-WAN capability and the genuinely independent infrastructure options in your area are mapped together. After installation the switchover tests are documented, and for customers under a maintenance agreement the line health goes into continuous monitoring — more often than not, we report that you have dropped to the backup line before anyone in the office has noticed.

FAQ