p2psp team mailing list archive

[Vicente Gonzalez <vicente.gonzalez.ruiz@xxxxxxxxx>]

2015-03-20 9:59 GMT+01:00 Ilshat Shakirov <im.shakirov@xxxxxxxxx>:

I don't know the procedure you are using to add peers to a team (how the
>> peers contact the splitter, receives the list of peers of the splitter and
>> send the [hello] message to the peers that are in the list), but I suppose
>> that this procedure is similar to one used in a real team.
>
> Yes, the peers are added to team as it described in DBS of rules (new peer
> sends HELLO to splitter, splitter sends list of peers, peer sends HELLO to
> rest of team).
>

OK.


> I have new hypothesis - how much time is spent for sending chunks? and
> for sending simple messages, like HELLO/GOODBYE. In peersim implementation
> there is no difference in times between these two types of message. If it's
> able to send 2 or more simple messages while one chunk is sending, then
> simulation results will match with expected results.
>

When a peer is introducing himself to the rest of the team only sends
[hello]s. To interlace [hello]s and chunks, the splitter should send first
to it a chunk, and this is done after receiving the list of peers from the
splitter. This does not happen.

In my opinion, a better result (not only for this experiment) could be
obtained if you assign a latency to each message that depends on the
message's length.


> This means that if the peer 34 were the trusted peer, the attacker would
>> send only one poisoned chunk before it is reported. For this reason, I
>> think that the order of the peers in the lists should have some influence
>> in the average number of poisoned chunks.
>
> There is such results already. I.e. cell D43 in google sheet
> <https://docs.google.com/spreadsheets/d/1pc6yb87xJy8gNkWSWvvCvjAjR6WBFdOzCbPvO-zEooU/edit#gid=0>.
>
>

I see :-)


>
> Also, I can remove the latency for simple messages (like hello and
> goodbye), and (I suppose) results will match with expected results.
>

No. Please, consider first my idea (longer messages will take more time
traveling that shorter ones).  And, randomize, if possible, the order (or
time) in which peers join the team.


>
> An accurate P2PSP simulator is very interesting for researching the
>> performance of the P2PSP. Any kind of advance/facility in this direction is
>> really appreciated :-)
>
>

> Ok, I will begin to write readme file about performing experiments =)
>

Again, thanks. Your work will be very helpful!!

Best,
Vi.

-- 
Vicente González Ruiz
Depto de Informática
Escuela Técnica Superior de Ingeniería
Universidad de Almería

Carretera Sacramento S/N
04120, La Cañada de San Urbano
Almería, España

e-mail: vruiz@xxxxxx
http://www.ual.es/~vruiz
tel: +34 950 015711
fax: +34 950 015486