keepcalculator

06 Numbers and Powers

Ethernet Link Aggregation vs Failover: Key Differences

Card No. 06·80419 · Updated · Standard formula, shown below

Ethernet link aggregation combines multiple physical network links into one logical link to increase bandwidth and provide load balancing, while failover uses a secondary link that activates only when the primary link fails, prioritizing availability over speed. In short, link aggregation is about making a faster pipe, and failover is about keeping the connection alive when something breaks. This article explains how each works, their real-world use cases, and how to choose between them.

Background: Why Combine or Back Up Ethernet Links?

Ethernet links are the physical cables and switch ports that connect devices in a network. A single link has a fixed bandwidth—for example, 1 Gbps or 10 Gbps—and if it fails, the connection is lost. To address these two limitations, network engineers developed two distinct strategies:

  • Link aggregation (also called port trunking, NIC teaming, or bonding) bundles two or more physical links into a single logical link. The combined bandwidth is roughly the sum of the individual links, and traffic is distributed across them.
  • Failover (or redundancy) keeps a second link idle or partially active. If the primary link goes down, traffic automatically switches to the backup link, minimizing downtime.

Both approaches are implemented at the data link layer (Layer 2) and are transparent to applications, but they solve different problems. Aggregation assumes both links are healthy and wants to use them simultaneously; failover assumes one link may break and wants to be ready.

How Link Aggregation Works

Link aggregation uses a protocol to negotiate the bundling of ports. The most common standard is IEEE 802.3ad, often implemented as Link Aggregation Control Protocol (LACP). When you enable LACP on a switch and a server, the two devices exchange frames to confirm that the links can be aggregated. Once established, the logical interface appears as a single link with a higher bandwidth.

Traffic distribution is not done per packet but per flow (e.g., based on source/destination MAC addresses, IP addresses, or TCP/UDP ports). This ensures that packets of a single conversation stay in order. Because of this, a single large file transfer on one TCP connection will only use one physical link—aggregation shines when there are many concurrent flows, such as in a server with many users.

If one physical link in the bundle fails, the aggregation group automatically removes it from the bundle, and traffic is redistributed among the remaining links. This provides a form of redundancy, but it is not the same as failover: the total bandwidth drops, and there is no standby link.

How Failover Works

Failover relies on a primary link and a backup link. The backup can be a separate physical interface or a second aggregated bundle. The network interface card (NIC) teaming software or the switch monitors the health of the primary link—often using a protocol like Link Aggregation Control Protocol (LACP) or a simple link-state check. If the primary link goes down (cable unplugged, switch port failure, or even a remote switch failure), the backup link is activated within milliseconds to seconds.

Failover can be implemented in two ways:

  • Active/passive: The backup link is idle and only carries traffic when the primary fails. This is the classic failover model.
  • Active/active with failover: Both links carry traffic (like aggregation), but if one fails, the other takes over the full load. This is a hybrid approach that combines bandwidth and redundancy.

Failover is often used for critical connections like internet uplinks or database links where a brief outage is unacceptable. The cost is that the backup link is underutilized during normal operation, so you pay for hardware that is not fully used.

Key Differences: Bandwidth vs. Availability

The core difference is the goal:

  • Link aggregation maximizes throughput by using all links simultaneously. It assumes both links are up and wants to use them.
  • Failover maximizes uptime by keeping a spare link ready. It assumes one link may fail and wants to avoid downtime.

Another difference is how they handle a link failure. In aggregation, a failure reduces capacity but does not disconnect the logical link. In failover, a failure triggers a switch to the backup link, which may cause a brief interruption (typically less than a second) but maintains the same capacity.

Configuration also differs. Aggregation requires both ends (e.g., server and switch) to support the same aggregation protocol (usually LACP). Failover can be done with any two independent links, even from different switches, as long as the NIC teaming software supports it. This makes failover easier to set up in some cases.

When to Use Which: Real-World Examples

Choosing between aggregation and failover depends on your priorities:

  • Use link aggregation when: You have many users or applications that generate multiple concurrent connections, and you need more bandwidth than a single link provides. Examples: a file server for a small office, a virtualization host with many VMs, or a NAS device accessed by several computers.
  • Use failover when: You need high availability for a single critical connection, and bandwidth is not the bottleneck. Examples: a company’s internet gateway, a database server that must be reachable 24/7, or a remote office link to headquarters.
  • Use both (active/active with failover): When you need both bandwidth and redundancy, you can configure an aggregation group with two links and enable failover to a third link, or use two separate aggregation groups with failover between them. This is common in enterprise data centers.

A common misconception is that link aggregation always doubles throughput. In practice, a single TCP stream will not exceed the speed of one physical link. If your workload is mostly large file transfers over one connection, aggregation will not help; failover would be more beneficial.

Common Problems and Misconceptions

Several issues arise when implementing these technologies:

  • Misconfigured LACP: Both ends must agree on the aggregation mode. If one side is set to static (on) and the other to LACP, the link may not come up.
  • Hash collisions: Aggregation uses a hash of packet fields to choose a link. If many flows hash to the same link, one link may be overloaded while others are idle. Adjusting the hash algorithm (e.g., using IP+port instead of MAC) can help.
  • Failover detection delay: Some failover implementations rely on link-state only, which does not detect a remote switch failure. Using a health-check protocol (like a ping to a gateway) can speed up detection.
  • Vendor-specific behavior: Although standards exist, some vendors implement proprietary modes (e.g., Cisco’s EtherChannel vs. LACP). Always check compatibility.

Costs and Alternatives

Link aggregation requires additional switch ports and a switch that supports LACP (most managed switches do). Failover also requires an extra port and possibly an extra switch for redundancy. The cost is mainly hardware and configuration time, not software licenses.

Alternatives to these Layer 2 solutions include:

  • Routing protocols (e.g., BGP with multiple paths): For WAN links, you can use dynamic routing to failover between ISPs.
  • SD-WAN: Software-defined WAN can use multiple links simultaneously and failover intelligently at the application level.
  • Multipath TCP: A newer protocol that allows a single TCP connection to use multiple paths, but it requires support on both ends.

For most local network scenarios, link aggregation and failover remain the standard tools.

Frequently Asked Questions

Does link aggregation increase speed for a single download?

No. Because a single TCP flow is hashed to one physical link, a single download will not exceed the speed of one link. Aggregation only helps when there are many simultaneous flows.

Can I use failover without a switch that supports LACP?

Yes. Failover can be done with two independent links, even connected to different switches, as long as your NIC teaming software supports it. No special switch protocol is required.

What happens if one link fails in an aggregated group?

The aggregation group removes the failed link and continues operating with the remaining links. The total bandwidth decreases, but the logical link stays up.

Is link aggregation the same as NIC teaming?

NIC teaming is a broader term that includes both aggregation and failover. Link aggregation is one mode of NIC teaming; failover is another.

Which is better for a home NAS?

If you have multiple users accessing the NAS, link aggregation helps. If you only have one user at a time, failover is more useful to avoid downtime, but a single 1 Gbps link is often sufficient.

Can I combine link aggregation and failover?

Yes. You can create two aggregated bundles and set up failover between them, or use an active/active team with failover if your NIC teaming software supports it.

Related

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.

More cards in this drawer

The whole Numbers and Powers drawer →