I'm experimenting with CEF on a 2621. I have a fast PC on each FE port,\neach PC running a small tight program that opens a TCP socket and\nmeasures the throughput to/from the other PC. I wrote the code myself,\nit's very fast and is definitely not the limiting factor on speed (if\nthe two PC's are connected through just a hub, throughput is ~80Mbps).\n\nWhen the 2621 sits between them, throughput is dramatically lower...\n~10 Mbps in one direction and ~6600 Kbps in the other (with CEF\ndisabled). I know the 2600 platform cannot sustain wire speeds, but at\nleast this gives us a baseline.\n\nSo I enabled CEF, expecting at least **some** increase in speed. Not\nso. In one direction (input F0/0 to output F0/1) the speed is\nunchanged; in the other (input F0/1 to output F0/0) the speed drops\ndramatically depending upon what interface settings I use.\n\nFor example, with "ip route-cache flow" on both FE's the speed is\nroughly the same as the CEF-disabled rate (~6600 Kbps). However, add\n"ip route-cache cef" to both interfaces and the F0/1 to F0/0 speed\ndrops to 700 Kbps (a drop of nearly 90%!). Removing all "ip\nroute-cache" statements from both interfaces yields a middle ground of\nabout 4500 Kbps.\n\nMeanwhile, in all cases the F0/0 to F0/1 traffic continues to be\n~10Mbps, utterly unaffected by any CEF or route-cache settings.\n\nThe only "extra" thing this router is doing is NAT. F0/0 is the outside\ninterface, and F0/1 is the inside interface. NAT is using only static\naddresses. I don't know if NAT would be the limiting factor on speed,\nor if that burden would be unidirectional. Does NAT force all packets\nto be process switched, effectively disabling CEF? I would think that\nCEF's routing table could still be built even allowing for NAT, just as\nit allows for ACL's.\n\nAny ideas gratefully accepted. I'm just trying to understand what's\nhappening here so I can get the most out of these routers.