Hi,
On Tue, May 03, 2011 at 09:48:03AM +0800, ?????? wrote:
> On Tue, May 3, 2011 at 2:18 AM, Sebastian Kiesel <[email protected]>wrote:
> > On Fri, Apr 29, 2011 at 03:25:55PM +0800, ?????? wrote:
> > > hi Sebastian ???
> > >
> > > - firstly, i agree that the creation of the map, which is tightly
> > related
> > > to the algorithm, is within the protocol.
> > > - secondly, i am not sure the information contained in current map
> > > structure is complete or not, because the current protocol describe
> > that :
> > >
> > >
> > > 1. the cost map just includes cost type and cost mode
> > > 2. the network map just includes PID and endpoint address
> > >
> > > does the information include the cell id or ap id?
> >
> > What would be your optimization strategy that requires exposing cell IDs
> > to the peers?
>
> in my understanding,currently, the peer in the wireless network, especially
> in the mobile network should be avoid be selected because of the limited
> network bandwidth and local capabilities, which should be included in ALTO
> information. what do you think?
this is a reasonable strategy. In GSM/UMTS-style networks, where all
data has to go all the way to the GGSN it is likely to be more efficient
to access a ressource "somewhere in the internet", even if the person
standing right next to you has a copy on his mobile.
In contrast, in a fixed network, the strategy might be to keep the
traffic local, i.e., peers connected to the same BRAS should better talk
to each other than to a distant peer via an expensive long-distance
link.
Both strategies can be implemented without using Cell IDs or DSLAM port
IDs in the ALTO client protocol. This is good, because a normal P2P
software probably does not know about such IDs. A normal P2P app only
knows IP addresses of other peers (e.g. learned from tracker or gossip)
and then has to decide whether to use it or not.
This is how we can hide the cell IDs, etc. from the P2P app: For example,
if you are an mobile + fixed operator, you could define some PIDs:
PID 1 = all the IP prefixes from which you give IP addresses to your
FTTH customers (look them up in your DHCP server configuration)
PID 2 = all the IP prefixes from which you give IP addresses to your
DSL customers (look them up in your BRAS / DHCP server
configuration)
PID 3 = all the IP prefixes from which you give IP addresses to your
mobile customers (not sure about that, but maybe look them
up in GGSN/NAS/HSS/?? configuration)
PID 4 = IP prefixes used by some parter ISP that has a special
arrangement with you for very cheap traffic exchange
(get the IP prefixes from BGP)
PID 5 = IP prefixes "behind" an expensive or constantly congested link
PID 0 = default, if IP address is not in any other PID
Then define some cost map that gives the relative cost, e.g.:
PID 0 = 100 (default cost)
PID 1 = 10
PID 2 = 50
PID 3 = 200
PID 4 = 20
PID 5 = 300
So for crating the maps you need special knowledge and methods fitting
your network technology (e.g. look into DHCP config, listen to BGP, ...).
For now, we do not want to standardize that.
The interface between ALTO server and ALTO client, which is used for
accessing the maps is in scope of the specifications. This rather simple
interface hides all the details of different access network technologies
(such as cell IDs) from the (peer-to-peer or other) ALTO clients.
Does this make sense to you?
Thanks
Sebastian
_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto
