Re: [BEHAVE] CGN REQ: Port Set Assignment

["Dan Wing" <[e-mail]>]

We need some common terminology.  Invent something.  

-d


> -----Original Message-----
> From: [e-mail]
> [mailto:[e-mail]]
> Sent: Thursday, March 17, 2011 12:25 AM
> To: Dan Wing; [e-mail]
> Cc: [e-mail]; [e-mail];
> 'Benson Schliesser'
> Subject: RE: [BEHAVE] CGN REQ: Port Set Assignment
> 
> Re-,
> 
> 
> 
> -----Message d'origine-----
> De : Dan Wing [mailto:[e-mail]]
> Envoyé : mercredi 16 mars 2011 20:34
> À : BOUCADAIR Mohamed OLNC/NAD/TIP; [e-mail]
> Cc : [e-mail]; draft-ietf-behave-lsn-
> [e-mail]; 'Benson Schliesser'
> Objet : RE: [BEHAVE] CGN REQ: Port Set Assignment
> 
> > -----Original Message-----
> > From: [e-mail] [mailto:[e-mail]] On
> > Behalf Of [e-mail]
> > Sent: Wednesday, March 16, 2011 7:38 AM
> > To: [e-mail]
> > Cc: 'behave' <([e-mail])>; draft-ietf-behave-lsn-
> > [e-mail]; Benson Schliesser
> > Subject: Re: [BEHAVE] CGN REQ: Port Set Assignment
> >
> > Hi Francis,
> >
> > By "random port set", I meant:
> >
> > * a set of ports which are randomly generated when a new outbound
> > packet is received by the CGN
> > * the port set is not contiguous
> > * the port set is not fixe
> > * for subsequent packet requiring a new port, a port from the pre-
> > allocated random port set is used
> > * when all the ports in the port set are in-use, a new random port
> set
> > is generated by the CGN, etc until reaching a quota.
> >
> > We can find a better name for this.
> 
> I described a continuum from completely random to completely fixed at:
>   http://www.ietf.org/mail-archive/web/behave/current/msg09182.html
> 
> which said:
> 
>    1. for every outgoing connection, create one mapping.
>    2. for an outgoing connection, create a "bin" of several mappings
>       using random public ports.  Subsequent outgoing connections will
>       use ports from the "bin".  When the "bin" is full, a new
>       connection causes a new bin to be created.  A bin is smaller or
>       equal to the user's maximum port limit.
>    3. Same as (2), but the ports allocated to a "bin" are consecutive
>       public ports.
> 
> Perhaps that is a good starting point.  Or perhaps this is better:
> 
>    1.  for a new outgoing flow, choose a random port and create a new
>        mapping.  There is a limit on the maximum number of ports.
> 
>    2.  for a new outgoing flow, choose several random ports and create
> a
>        "bin" of those ports.  This outgoing flow is assigned one of
>        those ports.  Subsequent outgoing flows will be assigned a port
>        from the "bin".  When the "bin" is full, a new flow causes this
>        process to be repeated (i.e., a new bin is created).  A bin is
>        smaller or equal to the user's maximum port limit.
> 
>    3.  Same as (2), but the ports allocated to a "bin" are consecutive
>        public ports.
> 
> Med: I don't know where to put the non-contiguous one single "bin" case
> (excerpt from (see http://tools.ietf.org/html/draft-bajko-
> pripaddrassign-02#section-4.1)
> 
>    "a non Continuous Port Range is assigned to a given customer's
>    device. In this example, the Port Range Value defines 128 Continuous
>    Port Ranges, each one with a length of 16 port values.  Note that
>    the two first Port Ranges are both in the well-known ports span
>    (i.e. 0-1023) but these two ranges are not adjacent.
> 
>    The following Port Range Mask and Port Range Value are conveyed in
>    DHCP messages:
>         - Port Range Value : 0000000001010000 (80)
>         - Port Range Mask : 0000000111110000 (496)
> 
>    This means that the 128 following Continuous Port Ranges are
>    assigned to the same device:
>         - from 80 to 95
>         - from 592 to 607
>         - ...
>         - from 65104 to 65119"
> 
> 
>    4.  Same as (3), but only one "bin" is allocated.