Re: [Din] Hyperledger evaluation on a mesh network

[Jehan Tremback <[e-mail]>]

Thanks Leandro. Our protocol does indeed leave pricing considerations
un-analyzed. This is something we will unfortunately not be testing
in a simulation like Jon, we'll have to see how things evolve in the
real world. Some of the insights gained from that paper may prove to
be useful.
The paper that Arjuna shared does indeed provide interesting and
rigorous experimental results on running a very small consensus
network in some specific network conditions. My criticisms were on the
"why" of doing that and do not detract from the quality of the paper.
Thanks for sharing.
--
  Jehan Tremback
  [e-mail]



On Wed, Apr 4, 2018, at 4:08 PM, Leandro Navarro wrote:
> The paper that Arjuna shared looks to the general topic of this WG
> “Decentralized Internet Infrastructure”, with an evaluation of the
> Hyperledger framework when running on a edge scenario of small Debian
> servers spread in a mesh network (not in the routers), instead of a
> typical data center env.> 
> 
>> On 4 Apr 2018, at 17:58, Jehan Tremback <[e-mail]> wrote:>> 
>> Even simpler, by adding a monetary price metric to a distance vector
>> protocol (we are currently testing this in production). Skimming your
>> paper, it looks like you are thinking along the same lines.> 
> Simpler but not too simple regarding pricing. Jon’s paper looked in
> 2010 at pricing related to congestion, fairness, incentives for
> cooperation, sending rates in flows and overall stability in a local
> mobile ad-hoc net. Still relevant work 8 years later.> 
>> Back to the Arjuna's post, the use of a blockchain implies that there
>> is some value to having an immutable transaction ordering mechanism.
>> In our protocol we conceptualize these ordered transactions as
>> "payments", while in Arjuna's paper the actual use of this
>> transaction ordering is left unspecified.> 
> … not an objetive of that research paper (evaluation of Hyperledger)
> != whitepaper.> 
>> Running the transaction ordering consensus protocol on the network
>> nodes themselves seems like a bad idea.>> These nodes have more of an incentive to mess with the ordering than
>> some faraway validators who know nothing about the specific
>> application and are only incentivized to order transactions
>> correctly. Also, the fact that there are always going to be many
>> fewer validators available on a local network means that the
>> consensus pool is smaller and more vulnerable to manipulation.> 
> Hyperledger has strong limitations when running in small servers,
> let’s forget about routers, but just relying on remote servers for
> local transactions can have drawbacks too.> 
>> I say, leave the transaction ordering to a global network of
>> validators who specialize in transaction ordering and leave the
>> networking to network hardware equipped with light clients. With
>> Althea, we are able to run everything on commodity routers on OpenWRT> 
> Thanks for your comments, Leandro.
> 
>> .
>> 
>> --
>>   Jehan Tremback
>>   [e-mail]
>> 
>> 
>> 
>> On Wed, Apr 4, 2018, at 1:04 AM, Jon Crowcroft wrote:
>>> or a much simpler approach:
>>> https://hal.inria.fr/inria-00466747/document
>>> 
>>> On Wed, Apr 4, 2018 at 3:09 AM, Jehan Tremback
>>> <[e-mail]> wrote:>>>> __
>>>> Why run full nodes on your networking hardware? One could achieve
>>>> the same security characteristics (or better) by simply using light
>>>> clients of a public blockchain on the networking hardware.>>>> 
>>>> --
>>>>   Jehan Tremback
>>>>   [e-mail]
>>>> 
>>>> 
>>>> 
>>>> On Tue, Apr 3, 2018, at 4:44 AM, Arjuna Sathiaseelan wrote:
>>>>> we recently did an evaluation of the hyperledger fabric in a
>>>>> community wireless network within the famous guifi.net[1]..>>>>> 
>>>>> will be of interest https://arxiv.org/pdf/1804.00561.pdf
>>>>> 
>>>>> Regards
>>>>> 
>>>>> -- 
>>>>> 
>>>>> Arjuna Sathiaseelan
>>>>> University of Cambridge | Ammbr Research Labs
>>>>> Personal: http://www.cl.cam.ac.uk/~as2330/
>>>>> N4D Lab: http://www.cl.cam.ac.uk/~as2330/n4d
>>>>> _________________________________________________
>>>>> Din mailing list
>>>>> [e-mail]
>>>>> https://www.irtf.org/mailman/listinfo/din
>>>> 
>>>> 
>>>> _______________________________________________
>>>> Din mailing list
>>>> [e-mail]
>>>> https://www.irtf.org/mailman/listinfo/din
>>>> 
>> 
>> _______________________________________________
>> Din mailing list
>> [e-mail]
>> https://www.irtf.org/mailman/listinfo/din
> 
> --
> Leandro Navarro
> http://people.ac.upc.edu/leandro  http://dsg.ac.upc.edu


Links:

  1. http://guifi.net/