![]() How adjacencies know how long they should keep the adjacency with restarted brother? Right, graceful timer. Meanwhile we can observe that in case of graceful restart control plane is restarted but data plane keeps table and going on forward packets. During GR the main router (Restarter) is restarted and all its adjacency routers (Helpers) keep the adjacency link state stable that means that no topology change will not occur. What is the main idea of GR? The main idea of GR to minimize the effect of sudden router restart which can cause (especially in OSPF and ISIS) topology change, adjacency change and SPF new calculation which will not help for faster convergence. It means that during the permanent forwarding no traffic lost will be occurred. It is not misspell- its forwarding not routing. But the concept remains the same: GR provides uninterruptable service of packet forwarding. As above mentioned FRR, GR also can used as OSPF GR, ISIS GR, BGP GR. ![]() The time of switchover takes nearly 50 ms as well as BFD.ģ) Graceful Restart (Non-Stop Forwarding/NSF). Ideally the concept works as following: BFD detects the failure and instructs IP FRR to launch the switchover to the backup route. FS doesn't install alternate path to FIB it only installs it to RIB! Get it? That's why even though we have FS in EIGRP, in the latest releases of EIGRP enhancements we have EIGRP FRR. And here we have the difference from EIGRP Feasible Successor mechanism. Fast Reroute performs Loop-Free Alternate path calculation and installs "repair routes" to the routing table and FIB. ![]() But MPLS, BGP, OSPF and IS-IS don't have this feature that's why they should use FRR to feel stable and confident. It means that in case of successor failure feasible successor keeps the situation in hands and we won't have packet loss. Because of that terrific feature EIGRP have preinstalled backup path in case of some network failure with primary route. All of us know that EIGRP has the concept of feasible successor during route calculation. CCNPs can claim "EIGRP can manage it without FRRs!" and will be right. This toll can be mentioned in several forms like IP FRR, OSPF FRR, IS-IS FRR, MPLS TE FRR but the sense of it is kept same: this protocol permits to switchover the route to the alternate loop free path in case of failure of primary path. That's why it should work in cooperation with other mechanisms which allows us not only detect but also fix the problem.Ģ) Fast Reroute (IP FRR). But we shouldn't forget that BFD is just DETECTOR of failure not TROUBLERSOLVER. BFD is used like additional tool and helper for routing protocols which allows us to manage our network failures more strictly. It means that end consumer barely notes some network interruption. Meanwhile BFD can detect link failure in 50 ms. 8 second until link failure will be detected, router creates new topology adjacency and SFP algorithm will be reapplied!! It's really uncool. As you know from foundation topics that in order to detect some link failure in network by means of routing protocol itself we need seconds, let it be OSPF with hello timer 2 seconds, and dead timer- 8 second. Many skilled readers can ask: why we need additional mechanism of link fault detection when we have hellos, dead timers, and others. In this topic we'll just "touch" concept each of them but detailed principle of work will be reviewed in coming chapters of RS Ninja.ġ) BFD (Bidirectional Fault Detection)- this mechanism designed to detect link failure in extremely short period. ![]() In view of this RFC and IETF created several of mechanisms which can defend our network from flapping and keep it stable. ![]() In case of distance-vector protocol its not so critical, but in case of link-state protocols it really does matter because all routers in area should recalculate their topologies according to topology change in network. Why? Cause any moment our network can be collapsed by unpredictable network change. But anyway we're still unhappy and unsatisfied with our architecture. It means that at least one IGP routing protocol has been integrated into our network. So Guys, our topics become more specific and complicated, it means that we're obtaining slowly ninja level of concept and can analyze our network performance more professional. ![]()
0 Comments
Leave a Reply. |