[GLLUG] traceroute MPLS messages

[James Courtier-Dutton]

On 22 March 2013 10:37, Alain Williams <addw at phcomp.co.uk> wrote:
> Yesterday evening my ISP (EntaNet) had problems and I lost connectivity for 6+ hours.
>
> I was away (Floss UK conference in Newcastle -- which was good BTW) and logged
> into my bytemark box to see what was up. I could not ping home & tried a
> traceroute, the output is below.
>
> I am not a networking wizard, would someone be able to explain (simply) what the
> MPLS lines are all about. The wikipedia article on MPLS gives me vague clue, but
> doesn't help with interpretation.
>
> Thanks in advance.
>
>
> $ traceroute mint
> traceroute to freshmint.phcomp.co.uk (78.32.209.33), 30 hops max, 38 byte packets
>  1  80-68-91-1.no-reverse-dns-set.bytemark.co.uk (80.68.91.1)  0.693 ms  0.560 ms  0.380 ms
>  2  89-16-188-38.no-reverse-dns-set.bytemark.co.uk (89.16.188.38)  0.828 ms  1.061 ms  1.564 ms
>  3  te3-3.cr02.man.bytemark.co.uk (91.223.58.78)  0.945 ms  0.881 ms  0.817 ms
>  4  gi2-25.man.core.enta.net (212.121.34.72)  0.962 ms  20.284 ms  20.137 ms
>  5  te5-1.telehouse-north0.core.enta.net (87.127.236.109)  7.700 ms  7.499 ms  8.569 ms
>      MPLS Label=9051 CoS=6 TTL=1 S=0
>  6  te5-2.gs1.core.enta.net (87.127.236.42)  7.513 ms  10.063 ms  8.835 ms
>      MPLS Label=388 CoS=6 TTL=1 S=0
>  7  te1-1.interxion.core.enta.net (87.127.236.86)  7.777 ms  7.653 ms  7.593 ms
>  8  * te2-3.interxion.dsl.enta.net (87.127.236.210)  8.220 ms *
>  9  * * *
>

Normally you would not see the MPLS label, because those points in the
network are not normally considered a router "hop".
Each "hop" decreases the TTL, and is used by traceroute to detect the "hops".
A long time ago, there was something called ATM.
ATM provided telecoms service providers a method to combine multiple
services over a single backbone architecture.
They could do "traffic engineering" on it, and guarentee bandwidths,
and simplified the clocking of the network by using plesiochronous
clocking.

ATM was good tried and tested reliable technology that might be in use
today if it was not for:
1) Fixed sized packets/cells. This is a compromise between small voice
packets, and large data packets.  MPLS has variable sized packets, so
does not need to compromise.
2) The manufactures of ATM equipment insisted on hanging onto their
high profit margins. Normally about 500%. MPLS equipment started at
the relatively cheap Ethernet switch price point. More like 100% or
below margin.
3) ATM had far better quality of service than MPLS did at the
beinning, but MPLS were happy to re-invent the wheel to eventually
improve MPLS to compare to ATM features. It tool MPLS equipment
development about 5 years before it was comparable to ATM features.
4) ATM could recover from equipment and cable faults in <50
milliseconds. MPLS/Ethernet could not, but is beginning to now.
You could switch off an ATM box in your backbone network, and a user
of a voice call would not even perceive the traffic re-routing
happening.

ATM and MPLS are "traffic engineering" protocols.
By this I mean that you can configure the network in such a way to
preserve QOS for particular services/customers even when the network
has failures.
This is different from a "routing" protocols network, that re-routes
on failure, but QOS is more of a best effort approach in case of
failures.

I personnally think that technically now days ATM and MPLS are
comparable, but ATM lost out due to costs and greedy manufactures
holding onto high profit margins. If ATM manufatures had dropped their
profit margins early on, I think ATM might have won the ATM vs MPLS
battle.