[Xorp-users] Performance test

Pavlin Radoslavov pavlin at icir.org
Sun Dec 2 10:56:19 PST 2007


José María Martín <manjon at terra.es> wrote:

> Adam, I know that but at this momment I can´t change my kernel version. 
> I am using a commercial secured OS
> based in Linux and the manufacturer has not update to 2.6 yet.

You should try at least to upgrade XORP, because 1.0 is quite old.
The current version is 1.4, but you might want to use even the
latest code from CVS, because it contains a number of bug fixes:

http://www.xorp.org/cvs.html

>From your email it wasn't clear to me whether multicast worked for
low bandwidth data. If it didn't then probably there is some setup
problem that needs to be fixed first.
If it worked for low bandwidth, did you have 100% losses for high
bandwidth?
One thing you should be aware is that the PIM-SM DR (Designated
Router) on the sender side might be using PIM-SM Register
encapsulation to unicast the data to the RP (where it is
decapsulated to native multicast). This adds more processing
overhead, and could explain the losses. Though, the threshold for
switching to native multicast is configurable (at the RP and the
last-hop PIM-SM routers on the receiver side) so this is one thing
to check in your configuration.

In any case, before you start playing with the XORP configuration
I'd highly recommend that you update its version.

Regards,
Pavlin


> thanks a lot Adam
> 
> Chema
> 
> 
> Adam Greenhalgh escribió:
> > I would recommend you upgrade your kernel version to one of the new
> > 2.6 kernels as there have been improvements in the kernel since 2.4.25
> > .
> >
> > adam
> >
> > On 01/12/2007, José María Martín <manjon at terra.es> wrote:
> >   
> >> First al all, thank you for the answers
> >>
> >> I am goint to do a data charge in a lab environment. The data will be
> >> multicast an unicast data flow.
> >> Xorp is running on a Linux 2.4.25 kernel
> >> I have different scenarios:
> >> 1. Static routing with xorp routing multicast (Pim-SM) being RP and not
> >> being RP. In the last case I will use bootstrap mechanism against two
> >> Nortel/Cisco routers.
> >> 2. OSPF routing (not xorp) and the same multicast environment.
> >>
> >> My hardware is intel xeon 2.8 ghz 1 gb ram and a gigabit ethernet
> >> network card. This server is running a checkpoint firewall too.
> >> In the first attempt the system crash ( crash = the source packets
> >> didn´t reach the destination machine) with 100.000 multicast packets
> >> flow. I am using 1.0 xorp version.
> >>
> >> Be run grateful for your opinions and ideas
> >>
> >> thanks in advance
> >>
> >> Chema
> >>
> >>
> >> Pavlin Radoslavov escribió:
> >>     
> >>>> Is there any performance test running over xorp? what is the throughput of xorp? Have you got any specifications about it?
> >>>>
> >>>> thanks in advance
> >>>>
> >>>> Chema
> >>>>
> >>>>         
> >>> Chema,
> >>>
> >>> Please clarify what kind of performance.
> >>> If you have forwarding performance in mind, this is defined by the
> >>> underlying system, because XORP is only the control plane.
> >>>
> >>> Regards,
> >>> Pavlin
> >>>
> >>> _______________________________________________
> >>> Xorp-users mailing list
> >>> Xorp-users at xorp.org
> >>> http://mailman.ICSI.Berkeley.EDU/mailman/listinfo/xorp-users
> >>>
> >>>
> >>>       
> >> _______________________________________________
> >> Xorp-users mailing list
> >> Xorp-users at xorp.org
> >> http://mailman.ICSI.Berkeley.EDU/mailman/listinfo/xorp-users
> >>
> >>     
> >
> >   
> 
> _______________________________________________
> Xorp-users mailing list
> Xorp-users at xorp.org
> http://mailman.ICSI.Berkeley.EDU/mailman/listinfo/xorp-users



More information about the Xorp-users mailing list