The MPLS-OPS Archive

["RamNarayan S" <ramnarayan.srinivasan@wipro.com>]

Title: Message

Hi,

 

Just a clarification on the statement regarding SONET frame size. Nothing to do with the original problem as such.

 

I am not sure from where the number 4470 comes in. The SONET basic frame size is 90 columns x 9 rows=810 bytes (for a STS-1 signal). Of these 3 columns go into SONET overheads (for management) and that leaves 87 rows x 9 bytes = 783 bytes for a payload. SONET also has the concept of concatenated payloads (OC-3c (155 Mbps), OC-12c (622 Mbps), OC-48c (2.5 Gbps) et al) where the payload need not be of the size of a basic STS-1 and could be fitted into a higher sized payload. E.g.. ESCON/FICON at 200 Mbps can be fitted into 4 STS-1s, but the nearest SONET concatenated group would be OC-12c (622 Mbps).

 

Also the L2 frames (w.r.t. Ethernet) as far as I know are not carried directly over the SONET frame. They are encapsulated inside a PPP frame (not necessarily a PPP session between the SONET nodes) or a GFP frame and then transported over SONET. ATM over SONET is another possibility. Anyway SONET doesn't care about what is transported through it. It is up to the higher layers to demarcate based on the application.

 

Some SONET vendor equipment have the option to have jumbo frames support enabled on their Ethernet interfaces that enable a L2 frame to have around 10,000 bytes, but typically these interfaces are not used in an MPLS n/w. Typically they would be used for Ethernet private lines and Ethernet VPNs over SONET.

 

RamNarayan S

-----Original Message-----
From: Tarun Bhalla [mailto:tarun.bhalla@wipro.com]
Sent: Saturday, March 22, 2003 3:34 AM
To: mpls-ops@mplsrc.com; Wulf Losee
Subject: Re: [MPLS-OPS]: MTU problems in EoMpls domain..

Wulf..

 

thanx for the prompt reply.. yeah!! i know

abt cisco/juniper providing certain cmds that

mention things like jumbo frames.. also in many

cases the core consists of Sonet.. where the frame

size is 4470 bytes.. so not a big problem there also..

 

bt as i already told this test is happening on some test

setups.. so for that we have 3 set of Linux machines

(Ingress, Egress and transit) and nothing more and

also having MPLS Linux patch to enable MPLS on them.

 

One option is that i change the NIC cards of the intermediate

interfaces to jumbo cards.. but am not going for that.. so have

to implement fragmentation there.. but that is also not a problem

rather my issue is since this problem is going to be there in

the MPLS networ, how has the community addressed this issue??

 

---[ Rest of the message deleted ]---