Welcome, Guest. Please login or register.

Welcome to the StellarVerse Forums!

I have created these forums as a *stable* and *lasting* alternative to other Stellar forums for developers and everybody else! Read more. Questions? PM -> mmathias! Thank you!

Author Topic: Paxos Algorithm  (Read 4020 times)

mmathias

  • Global Moderator
  • Newbie
  • *****
  • Posts: 37
    • View Profile
  • Tech Interests: C++, PHP, Backend
  • Twitter: @mmathias_mmint
Paxos Algorithm
« on: December 28, 2014, 08:34:56 PM »
Just found out about the Paxos algorithm (didn't have time to read the whole article though). Perhaps this could be provide some ideas for solving the consensus problem?

http://harry.me/blog/2014/12/27/neat-algorithms-paxos/?hn=1
¯\_(ツ)_/¯
Support StellarVerse: gLVXANgQtoNPaK9Nr4egdvh36jqJ9LMG1A

jed

  • Stellar Development Foundation
  • Newbie
  • *
  • Posts: 9
    • View Profile
  • Tech Interests: C++
Re: Paxos Algorithm
« Reply #1 on: January 03, 2015, 12:59:42 AM »
Yeah the new consensus algorithm has some ideas from Paxos.
Paxos only works when you trust all the machines (non-byzantine) running the network so it isn't a complete solution.

lab

  • Newbie
  • *
  • Posts: 7
    • View Profile
  • Tech Interests: C++, Scala
Re: Paxos Algorithm
« Reply #2 on: February 07, 2015, 03:48:43 AM »
I think  the new system has several similar concepts of google spanner http://research.google.com/archive/spanner.html .
but more open, compact and payment specific

xekoukou

  • Newbie
  • *
  • Posts: 8
    • View Profile
  • Tech Interests: Reactive model-oriented programming & mathematics
Re: Paxos Algorithm
« Reply #3 on: February 07, 2015, 04:44:33 PM »
Doesn't spanner require precise hardware clocks that are not available in common hardware. That is what I have read.

lab

  • Newbie
  • *
  • Posts: 7
    • View Profile
  • Tech Interests: C++, Scala
Re: Paxos Algorithm
« Reply #4 on: February 08, 2015, 04:47:36 AM »
xekoukou, you are right, they are assuming different circumstances, but the both are try to get similar balance between CAP,  high Availability via rigid time alignment.

IMHO, regarding transaction per second,

* not measured with tens of millions accounts.

i am quite interested in the design goal of new stellard. obviously only two choice:
between rippled and visa, beyond visa. there are conflict facts i find:
  • visa achieved max tps by mainframe (http://www.ibm.com/mainframe50/enginesofprogress/visa/) backed by tens of thousands talent engineer but SDF has less than ten.
  • stellar is designed from scratch without legacy processes(KYC, regulation, etc)
  • byzantium assumption sacrifices performence   

so what will be the answer?

xekoukou

  • Newbie
  • *
  • Posts: 8
    • View Profile
  • Tech Interests: Reactive model-oriented programming & mathematics
Re: Paxos Algorithm
« Reply #5 on: February 08, 2015, 12:20:25 PM »
I wish I knew the answer.

Given that stellar will need to perform transactions of several banks and multiple payment services in the future, I would guess that it will need to process more txs than visa.

I have a big interest in the algorithm myself. I wish we could get the paper beforehand.

lab

  • Newbie
  • *
  • Posts: 7
    • View Profile
  • Tech Interests: C++, Scala
Re: Paxos Algorithm
« Reply #6 on: April 15, 2015, 02:49:44 AM »
according to this news:
http://www.coindesk.com/stellar-founder-jed-mccaleb-new-protocol/

Quote
“From the beginning, we were trying to design it to be able to reach hundreds of millions of accounts, thousands of transactions. So we’ve stress tested to 100 million accounts and a few hundred transactions per second, and its holding up under those loads."