tracert vs. tracetcp

korbenda11as · ‎06-07-2010

Anyone who can explain to me the below tracetcp and tracert that I have gathered. From hop 4-12 show high latency but at the end it shows low latency result, is this acceptable? why at middle show high latency values? Thanks

1 1 ms 1 ms 1 ms 10.254.162.2

2 1 ms 1 ms 1 ms 10.254.168.29

3 51 ms 52 ms 50 ms 10.253.51.61

4 966 ms 1206 ms 1039 ms 10.202.57.55

5 1297 ms 1435 ms 1392 ms 10.193.80.7

6 951 ms 1113 ms 1036 ms 10.95.80.180

7 1093 ms 1023 ms 1201 ms 10.1.200.8

8 1169 ms 1308 ms 1077 ms 10.13.169.35

9 815 ms 644 ms 1034 ms 10.13.162.30

10 815 ms 727 ms 869 ms 170.249.184.20

11 589 ms 756 ms 669 ms 170.249.184.198

12 1045 ms 965 ms 748 ms 192.178.2.5

13 Destination Reached in 286 ms. Connection established to 199.200.25.190

Below is a simple diagram where tracetcp destination ip:443 from Network A to Network C was performed. A packet shaper is shaping the traffic from/to network A and from/to network C. Both Vlans traffic are shape in the shaper at network B.

B

MPLS

VLAN 1

VLAN 2

MAN

Packet Shaper

milan.kulik · ‎06-07-2010

Hi,

here is an explanation how ICMP packets travel through the MPLS backbone when running traceroute:

http://www.cisco.com/en/US/customer/tech/tk436/tk428/technologies_tech_note09186a008020a42a.shtml

(or http://www.cisco.com/application/pdf/paws/26585/mpls_traceroute.pdf for a .pdf version).

It shows why the result might be confusing showing long response time for routers in the middle of the path.

HTH,

Milan

korbenda11as · ‎06-08-2010

Hi Milan,

Thanks for your answer and the link you gave, however, is there any much easy explanation on this? Is the traceroute or tracetcp results shows acceptable/normal?

I did a lot of traceroute test that traverses MPLS clouds but this is the first time that I have this kind of results. Is the shaper has something to do with this? Any other simple explanation on this? Thanks,

milan.kulik · ‎06-08-2010

Hi,

the link I provided explains how the traceroute works through the MPLS cloud:

You either don't see the internal infrastructure (the second case in the document) or the response times are confusing as the ICMP packet is traversing the whole MPLS cloud back and force (the first case in the document and your case, too).

So in your case, the traceroute output reliable info is the path through the MPLS cloud and the number of hops.

The response times are misleading though (routers might also have some control plane policies configured to protect thier performance, see http://www.cisco.com/en/US/tech/tk648/tk361/technologies_tech_note09186a0080120f48.shtml#cpuimp) and the only important and reliable response time is that one from the target.

Any packet shaper on the path can also influence the response times depending on its configuration.

BR,

Milan