RARE user and assistance email list

[mc36 <>]

and please stay on-list, it makes things easier... once these things are 
agreed up,
and as i see similar things appear again and again, after a bit of time i 
think it's
easier that one just drops a thread url if someone asks a thing...
thanks


On 8/26/22 03:56, mc36 wrote:
> let me reverb it a bit
> in the eu part of the testbed, we obviously dont have layer1 connections
> that is, the bud-poz connection in phyisical looks like the following:
> bud---mx.bud--------------mx.poz---poz
> i would be surprised if they burnt a whole lambda for a testbed in the 
> undersea connections
> it's easy ti check, verify the pluggable, and that's maximum distance 
> capabilities in the datasheets
> in this topology, in the middle of the two mx-es, you have the whole geant 
> backbone
> as their bandwidth usage change, their systems automatically reroute the 
> xconnect
> that is, id their system goes crazy you can easily end up doing much longer 
> paths
> moreover, the mxes have huge buffers and geant have excessive transfers
> so please please use dedicated ports before writing any report about the 
> tofino capabilities
>
>
>
> On 8/26/22 03:34, mc36 wrote:
> > can we step back a bit?
> > what is the problem you're trying to solve?
> > i really don't issues you're facing
> > normally these characteristics are much easier could be acquired from the 
> > layer1 management systems
> > the throughput will be always there since these cannot reduce it
> > the delay, well, will depend as the labda will get rerouted over the fibres
> >
> >
> >
> > On 8/25/22 19:38, Marcos Felipe Schwarz wrote:
> > > I'd like to separate the different use cases:
> > >
> > > A. Setting Data Transfer Performance Expectations
> > > B. Troubleshooting - Locate performance issues
> > > C. Monitoring - Evaluate network behavior and alert on trend anomalies
> > > D. Stress tests
> > >
> > > perfSONAR is normaly used for A. and B.. It can aditionaly aid in C. in 
> > > conjunction with analysis tools.
> > > Regarding, bandwith tests. perfSONAR tries to use tools that mimic the end user experience. And since perfSONAR nodes are available on many places on the R&E networking 
> > > Community, it can provide checkpoits at the network to help to isolate in what region/conditions the problem manifests.
> > > The Tofino traffic generator would not fit weel with bandwidth tests on A., B. and C use cases. Since it is a stateless packet generator and does RFC2544 like tests. Where we 
> > > want to validate the maximum capacity of a device/system. But it won't provide a reliable user experience, since it doesn't implemente the Linux network stack or congestions 
> > > control, that would be presente on a traditional user's data transfer. It would also stress other elements in the system, so it should be only be used in controlled environments 
> > > (lab, maintenance windows...)
> > >
> > > I believe that having the GEANT P4 lab over dedicated links, would enable an interesting use case. Re-evaluate P4 dataplane changes in the field at scale. This would be PoC 1 
> > > and would fit in scenario D. Of course it would be safer to this kind of test in a lab.
> > > A second PoC would be to run tools in host OS of the device   and forward it through FreeRtr using CPU port, to achieve A. B and C. But we should consider the limitations 
> > > previously mentioned.
> > >
> > > Regards,
> > >
> > > Marcos Schwarz
> > >
> > > -----Mensagem original-----
> > > De: mc36 <>
> > > Enviada em: quinta-feira, 25 de agosto de 2022 11:19
> > > Para: Marcos Felipe Schwarz <>; Fr  d  ric LOUI 
> > > <>; Tim Chown <>; Ivana Golub 
> > > <>
> > > Cc: Xavier Jeannin <>
> > > Assunto: Re: RES: perfSONAR on RARE/FreeRtr
> > >
> > >
> > >
> > > On 8/25/22 15:56, mc36 wrote:
> > > > hi,
> > > >
> > > > On 8/25/22 15:41, Marcos Felipe Schwarz wrote:
> > > > > 1. supports all perfsonar test, including bandwith at 100G and above.
> > > > > With 2., we'll be limited by the CPU resources of the switch and 
> > > > > cpu-todataplane performance, which normally are very limited.
> > > > so true, moreover, if you do 100g measurements in every 10ms then the user 
> > > > traffic will be heavily suppressed...
> > > > imho doing a twamp measurements in every 10ms should give good enough 
> > > > results...
> > >
> > > btw, if there is no user traffic, then, the tofino have a very flexible 
> > > packet generator...
> > > i personally feel it a bit dangerous to have an api message to program it up, 
> > > but in between an ifdef...
> > > you know... why not? but building a 32*100g transmitter is imho illegal all 
> > > around the world...
> > >
> > > br,
> > > cs