Re: [tequila/sls] predefined services / SLS ?

From: Stefano Salsano (salsano@coritel.it)
Date: Wed Oct 18 2000 - 10:47:59 CEST

  • Next message: Marcus Brunner: "Re: [tequila/sls] predefined services / SLS ?"

    Carlos Alberto Kamienski wrote:
    >
    > > What is interesting is that we found the need to have pre-defined
    > > SLSs (we called it "Network Services" but the name can be misleading):
    > > for ease of negotiation and implementation only a subset
    > > of parameters can be specified, while the others are given by default
    > > with reference to a small set of "well known services".
    > > This is in perfect agreement with your idea that a well-known service
    > > identifer is just a compression of the SLS for ease of negotiation.
    >
    > Well-known services do negotiation easier, but I don't think they would
    > be just a compression of the SLS. I do not think an SLS will be always
    > sufficient.
    >
    > A quantitative well-known services would be a mix of:
    >
    > - explanation in a human language (eg, English)
    > - mathematics
    > - definition of parameters and possibly limit values for them
    >
    > My point is that one domain may not be able to figure out
    > services that other domains are willing to negociate just by observing
    > parameters in the SLS. I cannot give many examples, but the Virtual
    > Wire (VW) PDB is a good one! How could one domain know that one SLS
    > refers to VW and not to other different service (maybe not so
    > strict) using EF PHB ?

    Dear Carlos,
    I understand your concerns... I admit I have not the definitive answers
    for everything, but may be what you have described for a quantitative
    well known service may be expressed in terms of an SLS...

    Referring to your example, I see that the VW PDB could be expressed
    in an SLS with QoS parameters like:
    Absolute maximum delay: 100 ms
    Packet loss: 0 %

    A different service based on the EF PHB could have this parameters
    in the SLS (a quantitative one):
    99.9 delay percentile: 100ms
    Packet loss < 10^-6

    And a even different qualitative service based on EF could have:
    Delay: low
    Packet loss: very low
    (If an SLS also allows qualitative values :-)

    By the way, the user of the SLS could even be not interested in
    knowing what diffserv PHB is used in the diffserv network...

    Regards,
    Stefano

    -- 
    *******************************************************************
    Stefano Salsano
    CoRiTeL - Consorzio di Ricerca sulle Telecomunicazioni
                      
    E-mail  : salsano@coritel.it      URL     : http://www.coritel.it
    Tel.    : +39 06 20410029         Address : Via di Tor Vergata, 135     
    Fax.    : +39 06 20410037                   00133 Roma - ITALY          
    *******************************************************************
    



    This archive was generated by hypermail 2b29 : Wed Oct 18 2000 - 10:53:57 CEST