[Planetlab-devel] IPv6 support for MyPLC

[Bound, Jim]

All the production Linux and src pool implementations do Stateless Addr
Conf well its a no brainer.

/jim 

> -----Original Message-----
> From: devel-bounces at planet-lab.org 
> [mailto:devel-bounces at planet-lab.org] On Behalf Of Bound, Jim
> Sent: Thursday, November 09, 2006 12:49 PM
> To: Marc E. Fiuczynski; devel at lists.planet-lab.org
> Subject: RE: [Planetlab-devel] IPv6 support for MyPLC
> 
> Hi Marc,
> 
> OK I must be dreaming and don't have the time now to chase 
> all the mails.  But I thought I heard someone say each host 
> had multiple IPv4 addresses.  Now that you state this suggest 
> PLL simply use stateless addr conf and get the IPv6 host 
> address from a range of routers in the market that can do 
> this.  As you well pointed out the next step is putting in 
> the DNS but that is part II.
> 
> thanks
> /jim 
> 
> > -----Original Message-----
> > From: Marc E. Fiuczynski [mailto:mef at CS.Princeton.EDU]
> > Sent: Thursday, November 09, 2006 12:36 PM
> > To: Bound, Jim; devel at lists.planet-lab.org
> > Subject: RE: [Planetlab-devel] IPv6 support for MyPLC
> > 
> > Hi Jim,
> > 
> > Today all PlanetLab nodes are configured by the hosting site. 
> > Currently there is just one global IPv4 address per host. The 
> > technical contact at the hosting site assigns a static IPv4 address 
> > (either manually or via DHCP).
> > It is not clear we can assume that sites will bring up 
> DHCPv6 servers 
> > on our behalf (especially since we strongly request to be in a DMZ 
> > outside of their firewall).  Maybe they will, but a big 
> concern is the 
> > convergence timeframe
> > + the sheer amount of email we might have to send/receive 
> when asking 
> > + sites
> > to turn on IPv6 support on our behalf.  For this reason, 
> I'd like to 
> > arrive at a solution that factors out most (ideally
> > all) human interaction with sites.  The basic assumption is that a 
> > site supports IPv6 on their router and advertise global 
> IPv6 prefixes 
> > to the nodes.
> > 
> > The challenge is to also assign global IPv6 addresses to 
> each "virtual 
> > machine" (sliver) on each node.  Where do those addresses 
> come from, 
> > considering we wont have DHCPv6? Hence the discussion of trying to 
> > come up with an autoconf scheme that assigns a EUI per sliver.
> > 
> > > I believe stateless autoconfig will work ...
> > 
> > Which one?
> > 
> > > ... the problem is that
> > > any IP architecture view assumes some entity absorbs the 
> IP address 
> > > ...
> > 
> > I don't understand the problem you are stating.  Please 
> rephrase this.
> > 
> > > I had indirectly suggested a software VLAN switch to
> > multiple vservers
> > > on a single node for IPv6 with use of prefix bits and same EUI.
> > 
> > I think I understood this suggestion, but maybe I did not.  
> > Please give an explicit example.  Also, what would we need 
> to ask each 
> > site to make such a solution work?  Do they have to give us 
> a set of 
> > prefixes?  Is it reasonable to ask for dozens or hundreds of those 
> > prefixes per node?  How would one coordinate the prefix assignment 
> > amongst hosts at a site?
> > 
> > > IPv6 has done its job the tools exist.
> > 
> > No doubt!  I am just trying to figure how to make it work 
> in the real 
> > world when faced with ~100 "virtual machiens" per phyiscal host and 
> > the lack of a
> > DHCPv6 server.
> > 
> > Thank you for your feedback!
> > 
> > Best regards,
> > Marc
> > 
> > 
> 
> _______________________________________________
> Devel mailing list
> Devel at lists.planet-lab.org
> https://lists.planet-lab.org/mailman/listinfo/devel
>