pctek3050
asked on
Router Interface Errors and Collisions - Traffic Behavior on Highly Utilized 10Mb Metro Ethernet
I have a remote office with 40 users. They are currently sending all VoIP traffic and all data/Internet traffic straight to our headquarters via a VPN over 4 bonded T1s. The Internet traffic goes out the DS3 at headquarters and we host a Cisco VoIP phone system at headquarters.
We just had a 10 Mb Metro Ethernet line installed between the remote office and headquarters. I originally connected to two sites switch-to-switch but ATT called and said we could have no more than 50 MACs going across. Next I set up a Cisco 2621XM on each end and had the users up with data and voice. However, I had a couple voice quality complaints with my 4 test users and noticed collisions on my routers' outside interfaces. I setup the routers dmarc-to-dmarc and still had collisions when pushing large files between laptops connected to the routers. I also did several ping tests of all sizes between the routers with NO collisions.
Now I am setup laptop-to-laptop at each of the sites' dmarcs. The laptop NIC auto negotiated at 100 Mb full. I am pushing a 50 MB file across and monitoring the traffic with Wireshark. The file speed is approx 8 Mb/sec per the FTP app. Wireshark keeps showing "TCP dup ACK".
I also found an old 10 base T hub and connected the laptops straight into it. The NICs connected at 10 Mb. I got approx the same speed (8-9 Mb/sec) BUT &.NO TCP dup ACKs.
Is this normal behavior of high utilization on a Metro Ethernet link? Is this the cause of my collisions on the router interfaces? Would it help if my service provider connected to my routers' interfaces at 10 Mb? (I tried to force a 10 Mb connection on the router interface but it won't connect). Of course I would love to bump up my speed to 100 Mb Metro Ethernet, but it's much more expensive! I will of course do QoS for VoIP, but that doesn't do me any good if I am consistently racking up Interface errors/collisions.
My next step for troubleshooting is setting up the routers with my 10BaseT hub connecting the outside interfaces. I am thinking I will NOT show collisions or errors on the interfaces after pushing data between the connected laptops. If I am correct, do I have a case with my service provider?
We just had a 10 Mb Metro Ethernet line installed between the remote office and headquarters. I originally connected to two sites switch-to-switch but ATT called and said we could have no more than 50 MACs going across. Next I set up a Cisco 2621XM on each end and had the users up with data and voice. However, I had a couple voice quality complaints with my 4 test users and noticed collisions on my routers' outside interfaces. I setup the routers dmarc-to-dmarc and still had collisions when pushing large files between laptops connected to the routers. I also did several ping tests of all sizes between the routers with NO collisions.
Now I am setup laptop-to-laptop at each of the sites' dmarcs. The laptop NIC auto negotiated at 100 Mb full. I am pushing a 50 MB file across and monitoring the traffic with Wireshark. The file speed is approx 8 Mb/sec per the FTP app. Wireshark keeps showing "TCP dup ACK".
I also found an old 10 base T hub and connected the laptops straight into it. The NICs connected at 10 Mb. I got approx the same speed (8-9 Mb/sec) BUT &.NO TCP dup ACKs.
Is this normal behavior of high utilization on a Metro Ethernet link? Is this the cause of my collisions on the router interfaces? Would it help if my service provider connected to my routers' interfaces at 10 Mb? (I tried to force a 10 Mb connection on the router interface but it won't connect). Of course I would love to bump up my speed to 100 Mb Metro Ethernet, but it's much more expensive! I will of course do QoS for VoIP, but that doesn't do me any good if I am consistently racking up Interface errors/collisions.
My next step for troubleshooting is setting up the routers with my 10BaseT hub connecting the outside interfaces. I am thinking I will NOT show collisions or errors on the interfaces after pushing data between the connected laptops. If I am correct, do I have a case with my service provider?
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
um... well... on a note about the 50 mac's going across you could setup an IPsec tunnel between your routers and return to switch-connected operation.
GRE over IPsec (point to multipoint) introduces approx 9% overhead and IPsec (point to point, IP ONLY) introduces about 4%. However, by doing this you can 'hide' all of your site to site traffic (and unique MAC's) from your ISP and permit 'sanitized' data transfer over the WAN link.
Now, for the relevent part,
If you do this, you can create multiple parallel tunnels and separate by protocol/port (or anything else you can filter by an extended ACL). This would allow you to put wireshark (or just debugging mode) into each VLAN and look for collisions on a filtered basis and hopefully narrow down WTF is causing these collisions.
GRE over IPsec (point to multipoint) introduces approx 9% overhead and IPsec (point to point, IP ONLY) introduces about 4%. However, by doing this you can 'hide' all of your site to site traffic (and unique MAC's) from your ISP and permit 'sanitized' data transfer over the WAN link.
Now, for the relevent part,
If you do this, you can create multiple parallel tunnels and separate by protocol/port (or anything else you can filter by an extended ACL). This would allow you to put wireshark (or just debugging mode) into each VLAN and look for collisions on a filtered basis and hopefully narrow down WTF is causing these collisions.
ASKER
It turns out the Sherminator was dead on: "Cisco devices do not auto-neg well with half-duplex devices". I called my service provider and we hard-coded speed and duplex. So far no collisions on my routers' interfaces!
Thanks!
Jake
Thanks!
Jake
good to know; thx for letting us know the result.
ASKER
I would say it was unlikely that my 8 test users maximized the link consistently. I did not check the load at the time. I am going to call ATT to get the Interfaces set to a static setting. Anxious to see if that's it...
Thanks