[c-nsp] Load-balancing T1s on a Cisco 7600

Rodney Dunn rodunn at cisco.com
Wed Jul 27 00:06:46 EDT 2005


Try it at the global config level.

You are under  a subinterface configuration.

The global change sets how the hardware forwards
packets (ie: default is a src/dst ip hash) over equal
cost links.

full uses the L4 ports as part of the hash so if you
have a lot of traffic between two of the same ip addresses
but running on different ports you can have them hash over
different links which results in better load distribution.

Rodney

 

On Tue, Jul 26, 2005 at 09:57:01PM -0500, Cory Ayers wrote:
> > -----Original Message-----
> > From: Patrick Coppinger [mailto:pcoppinger at corp.earthlink.net]
> > Sent: Tuesday, July 26, 2005 2:19 PM
> > To: Ian Cox; Cory Ayers; cisco-nsp at puck.nether.net
> > Subject: Re: [c-nsp] Load-balancing T1s on a Cisco 7600
> > 
> > 
> > in a similar situation we had to add both of the following:
> > 
> > mls ip cef load-sharing full
> > mls flow ip full
> 
> I am running the following versions of IOS and do not see an option for
> mls ip cef on either.
> 
> s72033-pk9sv-mz.122-18.SXD4.bin
> s72033-advipservicesk9_wan-mz.122-18.SXE2.bin 
> 
> Router(config-subif)#mls ?                
>   ip       ip keyword
>   netflow  netflow command keyword
>   rp       rp
> 
> Router(config-subif)#mls ip ?
>   multicast  multicast keyword
>   <cr>
> 
> Router(config-subif)#mls netflow ?
>   sampling  sampling keyword
> 
> I also looked this command up using the Command Lookup Tool and don't
> see it listed.  I do see mention of the command in 12.2(17)SXB Release
> Notes but it shows [ simple | optimized | full ] options with no
> explanation other than it was implemented.
> 
> Has this done an adequate job of load-balancing for you?  Does it
> balance per-packet or source-dest-ID flow hash?
> 
>  
> > At 12:10 PM 7/26/2005 -0700, Ian Cox wrote:
> > >At 01:47 PM 7/26/2005 -0500, Cory Ayers wrote:
> > > >We recently upgraded multiple Cisco 7200s serving as aggregation
> > routers
> > > >to Cisco 7606s.  We found that ip load-sharing per-packet with CEF
> was
> > > >not an option due to the distributed architecture of a 7600.  We
> > > >attempted to use MLPoFR and MLPoATM, but neither of these worked,
> so we
> > > >opened a TAC case.  We were told that multilink PPP isn't going to
> work
> > > >on the 7600 and were advised to have our telephone company bundle
> the
> > > >circuits for us and hand us a single PVC.
> > >
> > >MLPPP and MLFR is supported on channelized and clear channel T1/E1s.
> > >MLPPP over ATM and MLPPP over FR is only supported on  single link in
> > >the bundle for LFI purposes. These features are only supported on
> > >FlexWAN/Enhanced FlexWAN and SIP-200. They are not supported on the
> > >OSM-2OC12-ATM-MM+.
> > >
> > > >  This is not a valid option as
> > > >it would affect 500 end-sites and mean dealing with nearly 30
> different
> > > >telephone companies.  Our other option appears to be to remove the
> > Cisco
> > > >7600s that we recently purchased in favor of a non-distributed
> > > >architecture.
> > > >
> > > >Currently, we have both Frame-to-Frame and Frame-to-ATM terminating
> on
> > > >the 7600.  The Frame-to-Frame circuits terminate on a PA-T3+ housed
> in
> > > >an Enhanced FlexiWAN.  Multiple DS3 and OC-3 ATM circuits terminate
> on
> > > >an LS-1010 and are fed to the 7600 on an OSM-2OC12-ATM-MM+.
> > >
> > >If your doing frame to frame then you can utilize MLFR FRF.16 if it
> > >is FR at both ends.
> > >FlexWAN
> >
> >http://www/en/US/products/hw/routers/ps368/products_configuration_guide
> _c
> > hapter09186a00803f37a8.html#wp84223
> > >SIP-200
> >
> >http://www/en/US/products/hw/routers/ps368/module_installation_and_conf
> ig
> > uration_guides_book09186a00802109bf.html
> > >
> > > >Has anyone successfully configured Multilink PPP over Frame or ATM
> in
> > > >this scenario?
> > >
> > >Yes on FlexWAN, Enhanced FlexWAN and SIP-200, see the above caveat,
> > >it is only supported for LFI, not supported bundling links.
> > >
> > > >Is there a work-around for CEF load-sharing per-packet that will
> allow
> > a
> > > >single flow to utilize multiple T1 circuits?
> > >
> > >Is just a single flow, or is the traffic multiple flows between the
> > >same two IP addresses. If it multiple flows between the same two IP
> > >addresses then "mls ip cef lod-sharing full" can balance the flows
> > >based on the L4 information in the TCP/UDP headers.
> > >
> > >
> > >Ian
> > >
> > > >Does the Cisco 10000 (ESR) boast similar issues with load-balancing
> > > >multiple T1 circuits?
> > > >
> > > >The Cisco 7200 has been a work horse for us for many years, but
> lacks
> > > >the port density, processor, and Gigabit fabric.  We have upgraded
> to
> > an
> > > >NPE-G1 in some scenarios, but this doesn't address high density
> circuit
> > > >aggregation.  What other routers are people using to fill this
> need?
> > > >
> > > >Thanks!
> > > >~cayers
> > > >
> > > >_______________________________________________
> > > >cisco-nsp mailing list  cisco-nsp at puck.nether.net
> > > >https://puck.nether.net/mailman/listinfo/cisco-nsp
> > > >archive at http://puck.nether.net/pipermail/cisco-nsp/
> > >_______________________________________________
> > >cisco-nsp mailing list  cisco-nsp at puck.nether.net
> > >https://puck.nether.net/mailman/listinfo/cisco-nsp
> > >archive at http://puck.nether.net/pipermail/cisco-nsp/
> > 
> > Patrick Coppinger
> > CCIE #14298
> > Senior Network Engineer
> > EarthLink, Inc.
> > 
> 
> 
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/


More information about the cisco-nsp mailing list