Basic question on "cell mode MPLS" and "frame mode MPLS"

Unanswered Question
Nov 6th, 2009

Folks, the difference between MPLS frame mode vs MPLS cell mode is that the cell mode is applied in ATM (which operated with fixed size cells)? Then the frame mode term is related to the PDU "frame" which is how data is transmitted at layer 2. Is that all?

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (2 ratings)
marikakis Fri, 11/06/2009 - 11:25

In most cases you have frame mode MPLS. In this mode MPLS typically adds its own header information (MPLS shim header) between the Layer 2 header and the Layer 3 header.

In ATM networks you can run either frame mode MPLS or cell mode MPLS. In a typical ATM setup you have routers at endpoints and an ATM cloud between the routers. The ATM cloud consists of ATM switches. The difference between the modes has to do with the capability of the intermediate ATM switches to speak MPLS.

If the ATM switches don't speak MPLS, you can run frame mode MPLS between the routers at the level say of ATM AAL5 PDUs (instead of cells) and this is transparent to the ATM switches (the switches do not know you run MPLS over them). If the intermediate ATM switches do speak MPLS, then you could run cell mode MPLS with the MPLS label information being mapped to the VPI/VCI field of the ATM cells.

Giuseppe Larosa Sat, 11/07/2009 - 04:12

Hello Marlon,

in cell mode MPLS the external label is resolved in the ATM VPI/VCI.

from an ATM point of view the Label switch path is considered an SVC switched virtual circuit that is setup on demand (when traffic needs to be sent).

the ATM LVC is negotiated over the PVC where label binding happens.

so in cell mode we have:

a PVC for signalling and setting up dynamic VCs on demand

a range of VPI and VCI values that represent the possible vpi#, vci# used for a specific Label SVC.

for saving on label space that is small labels/SVCs are assigned on demand when traffic is needed to be sent to intended destination.

this is quite different from frame mode where the MPLS label is assigned in an unsolicited way: that is regardless of traffic needs.

in frame mode a device can store received labels even if not used (not on current best path) this is called liberal retention.

just to give some figures:

in frame mode typical label space can be 1000000 of labels

in cell mode depending on the settings vpi and vci ranges cannot cover a so big space.

Cell mode has probably been historically the first implementation of MPLS but it is becoming rare.

signalling overhead is quite big for this on demand setting up SVCs that doesn't fit with topology driven CEF.

Hope to help



This Discussion