A second internet connection can improve resilience, but two circuits do not automatically provide independent paths or uninterrupted applications. Start by identifying the services that must keep working and the interruption each can tolerate.
Read the actual SLA
Availability percentages need a measurement window and a definition of downtime. For illustration, on a 365-day year:
| Availability | Theoretical annual downtime equivalent |
|---|---|
| 99.9% | 8 hours, 45 minutes, 36 seconds |
| 99.99% | 52 minutes, 33.6 seconds |
These figures are arithmetic equivalents: 365 days multiplied by the unavailable fraction. A monthly SLA, scheduled-maintenance exclusion or different measurement boundary changes how a contract is assessed.
Check the fault-reporting process, exclusions, restoration commitments and claim procedure. Mean time to repair (MTTR) is an average measure; it is not automatically a deadline for every incident. Remedies and liability depend on the agreement, so read the applicable terms before comparing offers.
Identify shared failure risks
Ask carriers about building entry points, local access infrastructure and upstream routing. Two brands may still depend on shared infrastructure. Also inspect the customer-side router, firewall, rack power and cable route.
A 4G or 5G backup can provide a useful alternative access medium, but it can share power, backhaul or carrier dependencies with the primary service. Check indoor signal, antenna placement, data allowance, addressing and performance during busy periods.
Record what is known and what the carriers have not confirmed. Do not label an unverified design fully independent.
Size the backup for priority work
A backup link may not carry the entire site's normal traffic. List essential applications and estimate their concurrent requirements. Define policies for large downloads, updates and other traffic during a failure.
Include hosted telephony, remote access and inbound services where applicable. An outbound browsing test alone does not establish that all business services will work through the backup connection.
Configure and test path selection
SD-WAN can use measured path conditions and policies to steer application traffic. Cisco's application-aware routing explanation describes measurements such as latency, loss and jitter. Cisco application-aware routing guidance.
Actual failover depends on detection thresholds, routing, addressing and the application session. Some sessions may reconnect. Measure the interruption rather than promising that users will never notice a change.
Test a disconnected primary link, degraded performance and restoration of the preferred path. Record application results, backup utilization, monitoring alerts and the actions expected from support staff.
BustanTech offers Cisco networking and managed services. Discuss a resilience assessment based on your carriers, applications and acceptable interruption.