JUNOS High Availability

Multivendor Architecture for High Availability

In Chapter 1, we described different redundancy models, including component, system, and site redundancy. While the discussion in that chapter implied that redundant pairs of devices should be identical, this condition was not stated outright. For example, when deploying a pair of switches to provide redundancy for a critical corporate local area network (LAN), the switches should at least be from the same vendor, and at best should be identical hardware running identical software. There are several reasons for this recommendation:

• Paired production network devices are expected to perform identically. Identical performance is best facilitated by identical platforms.

• The sparing scheme is simplified. A single collection of redundant components can be stored for the pair of devices.

• Interoperability testing burden is reduced. Identical devices are by default interoperable.

• Network management can focus on a layer in the scheme rather than devices within a layer.

For these reasons, when deploying products from an additional vendor, high availability is best protected by not mixing vendors’ products within a high availability pairing of devices.

Two Sensible Approaches

Since pairing devices from different vendors is not desirable, what is the best approach to deploying devices from multiple vendors in a production network? There are two sensible approaches.

Layered approach to multivendor networks

In the layered model, the network design team defines ...