bandwidth/performance ??? router 2621

[vince62s]

Config looks simple:

My ISP provides a router that I cannot control (a BINTEC) with a WAN (serial) 2 Mbps and a LAN interface.

If I test the bandwidth with a PC behind the Bintec, I have 2Mbps both ways -everything fine so far.

I connected a Cisco with two FaEth on the Bintec LAN.
I run a bandwidth test from a PC behind the Cisco: upstream is fine almost 1.8 Mbps and downstream is crap between 200K and 300K. I thought an IPsec Tunnel (with another site) was the pb but I removed it and same pb.

Router is 2621 with IOS 12.2(17).

ANY IDEA?

config is :

version 12.2
service timestamps debug uptime
service timestamps log uptime
service password-encryption
!
!
logging buffered 16000 debugging
enable password 7 094E471B11041C170201
!
memory-size iomem 15
clock timezone paris 1
ip subnet-zero
!
!
no ip domain-lookup
ip dhcp excluded-address 10.0.6.2 10.0.6.19
!
ip dhcp pool DDM
network 10.0.6.0 255.255.255.0
default-router 10.0.6.1
option 150 ip 10.0.5.10
dns-server A.A.A.A
netbios-node-type h-node
lease 2
!
!
!
call rsvp-sync
!
!
!
interface FastEthernet0/0
ip address x.x.x.x 255.255.255.248
ip nat outside
no ip route-cache
no ip mroute-cache
speed 100
full-duplex
no cdp enable
!
interface Serial0/0
no ip address
encapsulation frame-relay
shutdown
no arp frame-relay
!
interface Serial0/0.1 point-to-point
bandwidth 1024
shutdown
no cdp enable
frame-relay interface-dlci 108 IETF
!
interface FastEthernet0/1
description LAN
ip address 10.0.6.1 255.255.255.0
ip nat inside
speed auto
full-duplex
no cdp enable
!
ip nat pool branch x.x.x.x x.x.x.x netmask 255.255.255.248
ip nat inside source route-map nonat pool branch overload
ip classless
ip route 0.0.0.0 0.0.0.0 x.x.x.x+1
no ip http server
!
access-list 130 permit ip 10.0.6.0 0.0.0.255 any
no cdp run
route-map nonat permit 10
match ip address 130
!
!
dial-peer cor custom
!
!
!
!
!
line con 0
password 7 110B10171F1300090D27
login
speed 115200
line aux 0
password 7 110B10171F1300090D27
login
line vty 0 4
access-class 10 in
password 7 14151B190405212E2D25
login
!

 

[jneiberger]

Are you seeing any errors on the FastEthernet0/0 interface?

 

[vince62s]

no error at all

any clue?

 

[jneiberger]

I should have asked about the other interface, too. Do you see any errors on Fa0/1?

 

[vince62s]

nope.
no error at all.
The only strange thing is if I put Fa0/0 in auto nego it gets half duplex automatically, then no error.
If I force it in full-duplex, no error either.

looks like it comes from something else.

 

[jneiberger]

That is expected behavior if the interface it is connected to either does not support autonegotiation or has that function disabled for some reason. Since you're not seeing any errors, you're probably okay. If there were a duplex mismatch you would see input errors on the full duplex side and late collisions on the half duplex side.

On the other hand, it's interesting that you don't see any errors when it is in half duplex mode. Try setting it to AUTO and then running a bunch of traffic through it. You say that the Cisco router will fall back to half duplex. If the other side of that connection is really running at full duplex you should see a few late collisions on the Cisco router.

If you don't see any late collisions then the other device is most likely running in half duplex mode, in which case you might want to just leave it like that and try your test again to see if your throughput changes.

 

[jneiberger]

Also, make sure you have fast switching turned on by entering "ip route-cache" on your interfaces. If CEF is available, make sure it is turned on globally by entering "ip cef" in global configuration mode and then typing "ip route-cache cef" on each interface.

 

[vince62s]

what bothers me is that if it was a "cpu / routing (like cef)" issue then it would be a problem in both ways.

here the pb is just downstream.

 

[vince62s]

auto nego => half duplex => give errors.

I switched back to full duplex.
I added cef.

However I still have the same issue.
could it be a nat issue ?

this is driving me nut.

 

[jneiberger]

Are you seeing late collisions when you set it to auto? That probably means that the other side of that link isn't set to auto. You shouldn't have that problem if both sides are configured correctly.

I have no idea what the problem is, unfortunately. I'll ponder it a bit more and let you know if I think of something.

 

[nyingmeh]

Both ends of the link must be configutred to auto; fast switching turned on, otherwise this is strange to me.

 

[vince62s]

well I found without a good explanation.

I had a cross cable betwwen the 2 routers, no error.

I put a switch between the 2 router and now the speed is much higher..... go figure.
the bad one on this is that now I have some errors ( 1 per thousand ) I will go with it but very strange.