This article describes the issue of performance degradation when packets, sent across a high-end SRX device (SRX1400/3400/3600/5600/5800), need to go through the fabric link.

• Traffic from certain hosts may enter initially at node0.

•  Due to the network topology, the packet may be required to go across the fabric link (fab0 or fab1), in order to reach the final destination.

•  When this happens, it is observed that there is significant performance degradation when it goes through this path.

Traffic going across the fabric link (fab0 or fab1) is expected to see a performance drop of various degrees depending on the configuration and traffic pattern. For example, 25% drop is observed in some scenarios. For this reason, the fabric forwarding of packets needs to be used as a last resort. There are a number of reasons for this; but the following are key factors: 

• Traffic needs to be processed by both nodes
• There are extra traffic cycles that need to be executed when traffic comes over the fabric link; unlike traffic flow going through a single leg.
• When traffic goes over the fabric link, it is encapsulated with another header, which creates additional overhead in sending/receiving the fabric traffic.

19-11-2025: Minor, non-technical edit