[chord] Re: a question about chord simulator

[Ion Stoica]

Shu,

I believe that a trivial bound is that in the absence of leaves and 
joins, it
takes the system at most N^2 stabilization iterations to arrive in a state
in which each nodes know its correct successor (after each iteration at 
least
one node "improves" it's succesor, and there are N possible nodes in
the network). After that you need only log(N) other stabilization
periods to fix the fingers. David, correct me if I'm wrong.

However, in practice it will take much less time for each node to
learn its correct fingers. One option would be to periodically
check whether all nodes have correct fingers.

Hope this helps,

Ion

PS. For details you may take a look on the Chord TR at:
http://www.pdos.lcs.mit.edu/chord/

Shu Yanfeng wrote:

> Dear sir,
>
>  
>
>    Recently, I read the Chord paper and download the simulator. There 
> is one question puzzled me about the simulator: how do I know when the 
> network is stable, that is, all entries in finger tables are correct. 
> Is there some estimation about the waiting time after no node 
> joining/leaving the network. I want to know this, because I want to 
> compare lookup cost when network is consistent and inconsistent. Hope 
> to get your reply as soon as possible. Thank you very much!
>
>  
>
> Best regards,
>
> yanfeng
>