Ip-table lookup and mpls-label lookup

[networkquest]

Hi,

could u explain how mpls base lable-forwarding-lookup is faster than traditional ip-table lookup ?

I guess, in mpls - LSR needs to go through the lable-base to find right label agaisnt the FEC of the packet.

Also, in ip routing router needs to go through the routing-table.

both looks similar.

i like to know how mpls based lookup is faster than the traditional ip-lookup.

Thanks.

[smitesh kharecha]

Hi Robert,

In routing table, a route lookup is done first, depending on the lookup routing table then determine which interface it has to exit out and hence will prepare a L2 frame accordingly ( it can be Ethernet or FR or ATM, which ever is you WAN connection) which can be made in conjunction with Process switching, Fast Switching or CEF ( whichever you are running)

However in MPLS, lable table is made from CEF table. Hence lookup is at Layer 2 only.

So only one lookup in MPLS. Also, just in case if you are not much aware about CEF; CEF table is prepare before even a first packet is switched out of the device, as a matter of fact as soon as routing table is ready, your CEF table will be ready and packet will be ready to be switched out of the device....

Regards,

Smitesh

PS: Please rate helpful post...

[Peter Paluch · Accepted solution]

Smitesh,

I would personally put things slightly differently.

The principial difference between a lookup in the routing table and the MPLS LFIB is that the routing table lookup is concerned with longest prefix match, i.e. having potentially many (imprecise) matches and selecting the one that most closely resembles the destination IP address. On the other hand, the MPLS LFIB always performs lookups on fixed-length values and with equality operation, not with prefix-based logic. Hence, at least in theory, a routing table lookup is algorithmically more complex than a lookup in the LFIB, as finding a longest prefix match is more computationally intensive than simply finding a single matching value. Therefore the LFIB lookups should be faster.

Best regards,

Peter

[smitesh kharecha]

Peter,

Thanks for that interesting angle, however I'm still wondering what the difference is between my opinion and yours..

Maybe, just that I gave birds-eye view on L3 and L2 lookup and leaving operational difference to be understood while I said LFIB is derived from CEF...

Or is there something which I'm missing over here ??

Regards,

Smitesh

[networkquest]

Thank you Guys.

Am getting the point. It gives me some clarity over the discussion.

Could you people tell me how FEC plays its role in mpls and how it makes mpls better than ip-routing table lookup?

Thanks,

Robert

[smitesh kharecha]

Hi Robert,

Check this excellent write-up by Jeff Doyle:

http://www.networkworld.com/community/node/24093

HTH,

Smitesh

PS: Please rate helpful posts...

[medhat elsaeed]

Hi All,

if you checked the MPLS Fundamentals book, you will read that "if your sole reason for implementing MPLS in your network is to pursue the faster switching of packets through the network, it is a bogus reason "

because by the power of today's routers, it will be able to switch the IP packets as fast as switching the MPLS frames, because they are using the same CEF adjacency tables and the only difference is that the IP lookup is searching for 32 bit match and the MPLS is searching for 20 bit match and both lookups in the CEF FIB and LFIB tables which were already built

Regards