450 | High Availabilityw w w . d e l l . c o m | s u p p o r t . d e l l . c o mGraceful RestartGraceful Restart is supported on platforms: e c sGraceful restart (also called non-stop forwarding) is a protocol-based mechanism that preserves theforwarding table of the restarting router and its neighbors for a specified period to minimize the loss ofpackets. A graceful-restart router does not immediately assume that a neighbor is permanently down andso does not trigger a topology change. On E-Series, when you configure graceful restart, the system dropsno packets during an RPM failover for protocol-relevant destinations in the forwarding table, and istherefore called “hitless”. On the C-Series and S-Series, packet loss is non-zero, but trivial, and so is stillcalled hitless.FTOS supports graceful restart for the following protocols:• Border Gateway Protocol.• Open Shortest Path First.• Protocol Independent Multicast—Sparse Mode.• Intermediate System to Intermediate System.Software ResiliencyDuring normal operations FTOS monitors the health of both hardware and software components in thebackground to identify potential failures, even before these failures manifest.Runtime System Health CheckRuntime System Health Check is supported on platform: eFTOS runs a system health check to detect data transfer errors within the system. FTOS performs thecheck during normal operation by interspersing test frames among the data frames that carry user andsystem data. One such check is a data plane loopback test.There are some differences between the TeraScale and ExaScale line card and RPM testing:• The TeraScale card test contains a loopback from the RPM to the SFM and a loopback from the linecards to the SFM.• The ExaScale card test contains a loopback from the RPM to the SFM and a loopback from the linecards to the on-board TSF3.• For TeraScale, each line card and RPM periodically sends out test frames that loop back through theSFM. The loopback health check determines the overall status of the backplane and can identifies afaulty SFM. If three consecutive RPM loopbacks fail, then the software initiates a fault isolationprocedure that sequentially disables one SFM at a time and performs the same loopback test.
