3.7.2. Fault Tolerance and NIC Redundancy
💡 First Principle: Fault tolerance means the system continues operating when a component fails—ideally without any user impact. Every single point of failure in a system sets a ceiling on that system's availability. NIC redundancy eliminates the network as a single point of failure.
Fault Tolerance Levels
| Level | What's Redundant | Survives |
|---|---|---|
| Component redundancy | Individual parts (PSU, NIC, fan) | Single component failure |
| Server redundancy | Entire server (via clustering) | Server failure |
| Site redundancy | Entire data center (via DR) | Site failure |
NIC Teaming and Redundancy
NIC teaming (also called NIC bonding or link aggregation) combines multiple physical NICs into a single logical interface. This provides:
- Fault tolerance (failover mode): If one NIC or the connected switch port fails, traffic automatically moves to the other NIC
- Load balancing: Distributes traffic across multiple NICs to increase throughput
NIC Teaming Modes
The exam objectives put NIC teaming and load balancing under the same heading (redundant server network infrastructure) but list them as two separate things. Keep them apart:
| Mode | Description | Use Case |
|---|---|---|
| Failover | One NIC active, others standby | Pure redundancy, simple configuration |
| Link aggregation (802.3ad/LACP) | Switch-negotiated bonding, full bandwidth aggregation | Maximum throughput + redundancy; requires switch support |
Load Balancing
Beyond NIC-level load balancing, servers can sit behind load balancers (hardware or software) that distribute incoming client requests:
- Round robin: Each new connection goes to the next target in rotation. Simple but doesn't account for load — a named objective term.
- Most recently used (MRU): Keep using the path/target that most recently worked, and only move when it fails. The opposite philosophy to round robin: MRU optimises for stability (no unnecessary switching), round robin for even distribution. Best known as a VMware storage-multipathing policy, but CompTIA lists it as a general load-balancing method — a named objective term.
- Least connections: New connections go to the target with the fewest active connections. Better for variable-load workloads.
- IP hash: The same client IP always goes to the same server. Useful for session affinity.
- Hardware vs. software load balancers: Hardware load balancers (F5, Citrix ADC) have higher throughput and more features; software load balancers (HAProxy, nginx) are more cost-effective.
⚠️ Exam Trap: The objectives list round robin and most recently used (MRU) under load balancing, but failover and link aggregation under NIC teaming and redundancy. They are related but not interchangeable — a question naming MRU is asking about load balancing, not about how you bond two NICs.
⚠️ Exam Trap: Link aggregation (LACP/802.3ad) requires switch support and configuration. Simply plugging two NICs into a switch without configuring LACP does not create a team—it creates a loop. The switch must be configured to treat the connected ports as a port channel.
Reflection Question: A server with NIC teaming configured to distribute traffic across both members has one NIC connected to Switch A and another to Switch B. Switch A loses power. What happens to server connectivity, and what configuration would have made this more resilient?
