perfsonar development work

[Andreas Hanemann <>]

Hi Nicolas,

The outcome of the session in Sevilla (see also its minutes on the wiki, 
http://wiki.perfsonar.net/jra1-wiki/images/2/2d/Notes_closestMP.doc) has 
been as follows.

The identification of "closest MPs" has been considered as basis to 
investigate a given path. It is interesting to identify path segments 
that are shared between the path to be investigated and paths where 
regular measurements are running. For example, you can exclude some path 
segments from the problem search if you know that these are successfully 
used by the measurements. Therefore, a search should be conducted to 
automatically retrieve information about as many path segments as possible.

Currently, we are missing information about the actual paths where 
measurements are running. For example, for the measurements between 
HADES boxes there is currently no perfSONAR request to get to know the 
path of a current measurement. The outcome was that this has to be 
addressed by specific requests either for each kind of measurement or by 
the general implementation of something like a "traceroute MP".

For the functionality of investigating paths with the help of related 
MPs we have a separation of functionalities in mind. The visualisation 
tools (perfsonarUI) will have a logic to identify potentially related 
MPs in a first step and then check their measurements in a second steps. 
This functionality should be based on the Lookup Service which can 
answer the question whether a certain kind of MP is present at a given 
location. In addition, the visualisation tools need to have the 
possibility to check the neighborhood for which a request to retrieve 
the network neighborhood from cNIS should be implemented (I haved ask 
Marcin to change the requirement accordingly). The neighborhood should 
be differentiated between different kinds of neighborhoods (e.g. IP 
neighborhood, Ethernet neighborhood).

During the meeting Jeff mentioned that Aaron and Martin have been 
working in the past on a related document for an Internet2 Topology 
Service. Martin did not want to distribute the documentation at that 
point because it had some mistakes which may be misleading. He has not 
send a version yet (due to Supercomputing). After checking the document 
my idea is to more elaborate with Martin on the kind of queries that we 
need for the neighborhood and for the traceroute purpose.

I am sending this reply also to the perfSONAR dev list because more 
feedback to this non-trivial use of perfSONAR is appreciated.

Best regards
Andreas


Nicolas Simar schrieb:
> Hi Andreas,
>
> could you please summarise me what is the way that has been identified 
> for finding the closest MP along a path? The topic is growing in 
> importance and I'd like to get a clear understanding about it :-)
>
> Thanks a lot,


--
Dr. Andreas Hanemann
DFN-Verein
Stresemannstr. 78
10963 Berlin, Germany
Tel.:   ++49 30 884299 64
Fax:    ++49 30 884299 70
E-Mail: