Hi,
> Agree. However, if the IDs are not allocated in any
> hierarchy and they are randomly scattered across the DHT
> overlay, the lookup speed will be decreased heavily.
In general it has to be noted that in a DHT the number of lookup hops is a
function of the number of nodes participating in the DHT not a function of the
ID space.
> Of
> course, the tree-like mapping system like DNS or ALT has the
> same issue. So no matter DNS like mapping system or DHT, to
> make it scale better(in aspect of lookup speed) , the id
> should be allocated in some geographically aggregatable
> manner. Strict hierarchy is not flexible to support ID
> roaming since it may require some renumbering work, while the
> full flat structure is also not feasible. It's better to make
> some tradeoff.
I am not sure I follow you here. A geographic allocation of IDs would increase
the look-up speed, that is what you are saying right? What are the underlying
assumptions of this? If look-ups have an inherent locality property maybe. That
is, end-host communicate with higher probability with nodes that are
geographically close AND the node that is responisble for the lookup is in the
same geographic region AND the underlying network topology follows this
geographic constraints ONLY then this statement it true. Considering a popular
application such as BitTorrent which also opens many connections simulatneously
(i.e. would trigger numberous lookups), my personal experience is that locality
does not impact the peer choice an awful lot. To deal with non-locality today
we employ caches and CDNs (that's at least partly why we do it). And I think in
general it is safe to say that we will continue to do it.
Rolf
NEC Europe Limited | Registered Office: NEC House, 1 Victoria Road, London W3
6BL | Registered in England 2832014
--
to unsubscribe send a message to [EMAIL PROTECTED] with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg
