p2psp team mailing list archive

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

2014-02-15 20:46 GMT+01:00 Juan Álvaro Muñoz Naranjo <juanalvaro83@xxxxxxxxx
>:

>
>
>
> >1. Relax the rules that the peers and the splitter use to reject the
> >un-supportive peers.
>
> What do you have in mind? A higher threshold?
>
> Here's an idea to reward efficient peers: the faster a peer sends his
> packets, the higher he should be in the peer list of the rest of the team
> (I place good peers at the beginning of my peer list so they are rewarded
> by receiving my packets sooner). Thus, "fast" peers would be progressively
> climbing to the first positions of all peer lists in the team during a
> session.
>
> However, we have no way of measuring how fast packets arrive yet. Any
> suggestion?
>

Good idea!

I think that we could assign a reward to the peers in the list based on the
order they deliver their chunks. For example, if there are 10 peers, we
assign a reward to the first peer 10 that sends a chunk to us, 9 for the
second one, and so on. Next, we simply sort the list of peers using these
(cumulative) rewards.

If we want to avoid having to sort the list (which need not be too
expensive if we think that the list is sorted whenever a chunk is received
from the splitter and when the team grows, the frequency with which one
receives a chunk from the splitter decreases linearly with the size of the
team) could simply sort the list in the order in which they arrive the
chunk.

I think that this idea deserves to be included in the White Paper of the
P2PSP! Maybe in a specific set of rule. It could improve the performance of
the protocol and the computational cost is low.

Thanks!


> Juan
>
> --
> Enviado desde mi teléfono con K-9 Mail.
>



-- 
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