In IT Network Architecture, you’ll frequently hear terms such as path diversity, path redundancy, diverse paths, and physical path diversity used interchangeably. However, these terms don’t all mean the same thing.
Understanding the differences is particularly important when designing enterprise WANs, data center interconnects, cloud connectivity, and BGP routing architectures.
Although these concepts are closely related, each addresses a different aspect of network resiliency and availability.
Understanding Network Path Redundancy and Diversity
The following table explains the most commonly used terms in enterprise networking and provides practical examples.
| Term | Meaning | Example |
|---|---|---|
| Path Redundancy | Multiple network paths exist to maintain connectivity when a path fails. | Two independent BGP connections to AWS Direct Connect. |
| Path Diversity | Network paths use different infrastructure to reduce shared points of failure. | One cloud connection through New York and another through Chicago. |
| Diverse Paths | A general term describing connections that follow different network routes. | Two WAN circuits traversing separate network paths. |
| Physical Path Diversity | Connections use physically separated infrastructure, such as different fiber routes, conduits, or building entrances. | Two fiber circuits entering a data center from opposite sides using independent conduits. |
| Carrier Diversity | Multiple telecommunications providers are used to reduce dependency on a single carrier. | One WAN connection using Lumen and another using Verizon. |
| Geographic Diversity | Network connectivity is distributed across geographically separated facilities or regions. | Cloud connectivity through Equinix New York and Equinix Chicago. |
| Disjoint Paths | Paths are engineered not to share specified network resources, such as intermediate links or nodes. | Two paths engineered without common intermediate transport links. |
The Most Important Distinction: Redundancy Does Not Mean Diversity
One of the most common misconceptions in network architecture is assuming that redundant connections automatically provide path diversity.
Redundancy and diversity are not the same thing.
A network can have multiple routers, circuits, and BGP sessions while still depending on the same underlying physical infrastructure.
Scenario 1: Redundant but Not Physically Diverse
Consider an enterprise WAN design with the following characteristics:
- Two edge routers.
- Two separate WAN circuits.
- Two BGP sessions.
- Both circuits traverse the same physical fiber conduit.
At first glance, the design appears redundant. If one router or logical connection fails, the other can potentially maintain network connectivity.
However, if a construction crew damages the shared fiber conduit, both circuits could fail simultaneously.
The design provides logical redundancy but lacks physical path diversity.
This represents a shared physical failure point that may not be immediately apparent from a logical network diagram.
Scenario 2: Redundant and Diverse
Now consider a network architecture with the following characteristics:
- Two independent edge routers.
- Two separate transport connections.
- Connectivity through geographically separated data centers.
- Independent fiber routes and building entrances.
- Separate carrier infrastructure where available.
In this design, the network provides multiple layers of redundancy and diversity.
A failure affecting one router, transport circuit, or geographic location is less likely to impact the alternate network path.
However, geographic separation alone does not guarantee complete physical diversity. The underlying carrier infrastructure must still be evaluated for shared-risk dependencies.
Applying Path Diversity to Cloud Connectivity
These distinctions become particularly important when designing connectivity to public cloud providers such as AWS, Microsoft Azure, and Google Cloud.
For example, consider an enterprise architecture using AWS Direct Connect through two Equinix metropolitan locations:
- New York: Primary or active AWS Direct Connect connectivity.
- Chicago: Additional AWS Direct Connect connectivity for redundancy and geographic diversity.
An appropriate architectural description would be:
“We have redundant AWS Direct Connect connectivity with geographic diversity across our New York and Chicago Equinix metropolitan locations.”
This accurately describes the geographic separation of the connectivity.
However, claiming that the architecture provides fully physically diverse paths would require additional validation.
Network architects should verify the underlying infrastructure, including:
- Carrier fiber routes and transport facilities.
- Building entrances and physical conduits.
- Equinix cross-connects and network handoffs.
- Carrier points of presence (POPs).
- Shared physical infrastructure between different providers.
- Other potential common points of failure.
Even when two different telecommunications providers are used, their fiber circuits may share portions of the same physical infrastructure.
Carrier diversity does not automatically guarantee physical path diversity.
Understanding Shared Risk Link Groups (SRLG)
Another important concept in network resiliency is the Shared Risk Link Group (SRLG).
An SRLG identifies network links that share a common risk of failure because they depend on the same underlying resource or infrastructure.
For example, two separate WAN circuits may belong to the same shared-risk group if they traverse:
- The same underground fiber conduit.
- The same bridge or tunnel.
- The same physical fiber span.
- The same optical transport equipment.
- The same carrier facility or other critical shared infrastructure.
When designing highly available WAN and cloud connectivity, identifying SRLG dependencies helps reveal failure scenarios that traditional logical network diagrams may overlook.
Two circuits that appear independent from a routing perspective may still share a significant physical risk.
Recommended Terminology for Network Architecture Documentation
When preparing a High-Level Design (HLD), Low-Level Design (LLD), or network architecture review, consistent terminology helps communicate the actual resiliency capabilities of the design.
I recommend using the following terms:
- Path Redundancy: When describing alternate network paths and failover capability.
- Path Diversity: When describing paths that use different network infrastructure to reduce shared failure exposure.
- Geographic Diversity: When describing connectivity distributed across separate metropolitan areas, data centers, or cloud regions.
- Carrier Diversity: When multiple telecommunications providers support the connectivity architecture.
- Physical Path Diversity: When independent physical transport routes have been verified.
- Disjoint Paths: When specifying paths that must not share particular network links, nodes, or other resources.
Final Thoughts: Design for Resiliency, Not Just Redundancy
Network resiliency requires more than simply adding another router, circuit, or BGP session.
A properly designed architecture should consider multiple failure domains, including network devices, telecommunications providers, transport infrastructure, physical facilities, and geographic locations.
Redundancy provides alternate paths, while diversity reduces the likelihood that those paths will fail because of the same underlying event.
For network architecture documentation, I recommend using “Path Redundancy and Diversity” as a section heading, followed by a clear explanation of the specific layers of redundancy and diversity the design provides.
Bottom line: A network can be redundant without being truly diverse. Understanding and documenting the difference is essential when designing resilient enterprise WAN, data center, and cloud connectivity architectures.