Re: [RRG] Using a DHT to map identifiers onto locators

[Robin Whittle <rw@firstpr.com.au>]

Hi Laurent, Luigi and Olivier,

Regarding your LISP-DHT proposal:

http://inl.info.ucl.ac.be/publications/lisp-dht-towards-dht-map-identifiers-locators


For the benefit of people such as myself who are not familiar with
Distributed Hash Tables, it would be great if you could some
concrete examples of how your proposal would work.

For instance, if you choose some example figures for as large a
deployment as you think LISP would ever achieve:

  Number of distinct end-users.

  Number of EID prefixes - many end-users will have quite a few.

  Number of ITRs (They need to join in some sense, to be able to
                  make mapping requests and receive replies.
                  Or do you have some kind of concentrator for
                  queries from a bunch of nearby ITRs, in which
                  case this is a local query server, presumably
                  with cache?)

  Number of ETRs, or at least whatever the devices are which are
  authoritative query servers for mapping replies.

  Within these, perhaps assume two or three redundant servers for
  each EID, though of course some servers would be for multiple
  EIDs.

Now, could you describe the physical and logical structure of the ring?

Then, could you describe the sequence of events such as follows?  An
ITR in some arbitrary location places a query into the system, that
query is somehow sent around the ring to one of the authoritative
query servers (ETRs, in LISP, I understand) and how that query gets
back to the ITR.

Can the query come back straight via the Internet, or do you need it
go to via the ring as well?

This ring could be very big indeed, with hundreds of millions of
end-users, each with their own multiple authoritative mapping sources.

How fragile is it?

Do query (and response?) packets jump from one node to another on
the ring, or are they passed from one peer to the next?

To what extent does the communications traffic and computational
load of any one node depend on that of its neighbours?

To what extent does one node get to see, or be able to infer
something about, activity of other nodes?

How are multiple authoritative servers for the one EID accommodated,
with the queries being sent to one or all?  If all, how to avoid
them all sending back responses?

To what extent are PKI or other cryptographic techniques used, and
how in practice would trust, certificates etc. be administered?

I am opposed to any global query server system - CONS, ALT, your
modified DHT proposal or Bill Herrin's TRRP - so no matter what you
write I am unlikely to be impressed.

Still, I think other people on the list would be interested and I
think it is good to understand and critique all substantial proposals.


  - Robin               http://www.firstpr.com.au/ip/ivip/



--
to unsubscribe send a message to rrg-request@psg.com 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