perfsonar development work

[Joe Metzger <>]

On Jul 23, 2007, at 11:35 AM, Wenji Wu wrote:

> > Some OWD tools report the TTL. Is this sufficent?
>
> TTL just tells us that there are route changes. Still we need to  
> differentiate whether it is route flaps or load balancing cause the  
> route changes.
>
>
> > There are several known causes for persistent route flapping on short
> > timescales. There are techniques to dampen announcements caused by  
> > these
> > flaps but they are not all universally deployed.
>
> That is true. Some routers do not have the MRAI configured.

Flapping occurs at both the IGP and BGP levels. MRAI is not relevant to
a whole class of problems that introduce route flapping, especially with
protocols that are trying to achieve sub second, or sub 60 ms  
convergence.

> I just think about another hint:  route flaps usually cause  
> instantaneous disconnectivity, and lead to packet drops. If there  
> are no other reasons, load balancing should not lead to packet  
> drops. So we might combine the packet drops to differentiate the  
> route flaps and load balancing.

Not necessarily.  Routes can flap between 2 good paths with no packet  
loss.


> > It is not just dependent on cross traffic. For example the Juniper
> > load balancing algorithm can be configured to balance on various  
> > parts of the packet
> > header including IP address and TCP/UDP port numbers.  But in  
> > reality it doesn't look
> > at the IP addresses and port numbers, it looks at specific offsets  
> > in the header. The TCP
> > destination port and the ICMP checksum bytes are at the same  
> > offset. So ICMP based
> > traceroutes generate confusing results.
>
> here, my understanding is this: if the algorithm is configured to  
> balance on various parts of the packet header including IP address  
> and TCP/UDP port number, the load balancing is inter-flow based.  
> for specific flow (some src, dest, port), there is no load balance  
> at intra-flow level (or packet level) at all.

Provided the 'flow' is TCP or UDP yes. If the test is ICMP, then a  
'flow' description consists of
a series of packets with the same src IP, dst IP, Type, Code and  
Checksum.




> weji
> > > -----Original Message-----
> > > From: Joe Metzger [
> > > Sent: Friday, July 20, 2007 4:23 PM
> > > To: 
> > >
> > > Cc: 
> > >
> > > Subject: Question about interpreting One Way Delay data.
> > >
> > > Hi,
> > > An interesting question came up in the course of debugging a
> > > performance problem today.
> > > It is not real important from an operational perspective, but it is
> > > intriguing from
> > > an academic point of view.
> > >
> > > Is there a way to analyze one way latency measurements between
> > > endpoints over long
> > > slow congested paths and differentiate between route-flapping and
> > > load balancing?
> > >
> > > If so, do you need access to raw data, or can you do it from the
> > > type of
> > > summarization that the DFN and I2 tools are typically doing today?
> > >
> > > Thoughts?
> > >
> > > --Joe