p2psp team mailing list archive

[Ilshat Shakirov <im.shakirov@xxxxxxxxx>]

Hello!,

I just prepared the new pull request : https://github.com/P2PSP/sim/pull/3

Also, there is new result (with fix from previous mail) :
https://docs.google.com/spreadsheets/d/1pc6yb87xJy8gNkWSWvvCvjAjR6WBFdOzCbPvO-zEooU/edit#gid=506203193
(result3 tab)

As you can see, there is something strange in experiments with 50 peers,
and T>1. But now DIFF decreases with number of experiments (10 000 instead
of 1000 I have tested).

Also, I prepared guide in README file, now you can perform experiments of
your own, you need only JDK on local machine. You can change the number of
trusted peers in attackers in next lines:

> init.2.malicious_count 1
> init.2.trusted_count 1
>
Also, you can change the size of the network in the fist line.
And you can uncomment these lines to see all the information about network
(peer's buffers and neighbors)

> #control.0 sim.src.PeerObserver
> #control.0.protocol 2
> #control.0.step CYCLE*1
>
There is file sim/utils/average.py - it's a file for computing average
value of poisoned chunks from many experiments. You simply should call it
like 'python average.py -i log.txt'. Log.txt is a output from simulator.

That's all, next I will try to investigate the strange values from
simulator (with 50 peers).

Thanks!

2015-04-03 10:03 GMT+05:00 Ilshat Shakirov <im.shakirov@xxxxxxxxx>:

> Hello!,
>
> I found problem in the concept of simulation =)
> The malicious peer connects to the team after some delay and it's a
> problem. Because of that, i.e. with parameters N=9, T=6, A=3, splitter will
> give the peer list with first trusted peers for attackers. And the result
> will be 3. But expected is 3.4. And this is why the expected differs from
> result.
> I will fix it soon, and update the sheet. I think after this update the
> DIFF rows must become equal to 0.00 =)
>
> Thanks for feedback!
>
> 2015-04-03 3:49 GMT+06:00 Vicente Gonzalez <
> vicente.gonzalez.ruiz@xxxxxxxxx>:
>
>>
>>
>> On Thu, Apr 2, 2015 at 10:19 AM Ilshat Shakirov <im.shakirov@xxxxxxxxx>
>> wrote:
>>
>>> Hello!,
>>>
>>
>> Hello
>>
>>
>>>
>>> I began fill the results to table -
>>> https://docs.google.com/spreadsheets/d/1pc6yb87xJy8gNkWSWvvCvjAjR6WBFdOzCbPvO-zEooU/edit#gid=506203193
>>> (result3 tab)
>>> I think formula works, but Im a little bit confused about diifs
>>> 0,1-0,12. May be it's some bugs in stat gatherer module of simulator, maybe
>>> there some mistake with formula.
>>>
>>> Soon, I will prepare scripts for running simulation without ide. And I
>>> will attach log files (with each peer' buffers) to experiments.
>>>
>>> What do you think? =)
>>>
>>
>> Sorry, I don't know is there is a mistake somewhere (in any case, results
>> are quite good). But I have a question. Does DIFF decrease with the number
>> of experiments?
>>
>> Best,
>> Vi.
>>
>
>