[chord] question

[Emil Sit]

On Tue, 05 August 2008 at 15:15 (+0700), Herry Sitepu wrote:
> First i want to congratulate you for your Ph.D. I hope you don't mind if i
> ask question related with your work. I want to know is it possible to
> collect data such as failure rate for a time interval at DHT level (Chord,
> Accordion) from lsd process? Could you give me a hint how to do this? Thanks
> in advance.

Herry,

Of course it would be possible to instrument the code to
collect information you are interested in.  What do you mean
by a failure?  When a node individually directly detects that
an RPC it makes to a remote node has failed?

It may be that 'lsdctl rpcmstats' is sufficient to obtain that
particular statistic.  You may have to add some code to allow
the stats to be reset so that you can measure the particulars
you are interested in.  You can model that in theory after
how the rpcstats -r (not rpcmstats) command is handled.

The stats themselves are maintained in chord/comm.C (and
chord/stp_manager.C); the handling of the lsdctl RPCs is
done in lsd/lsd.C.

Feel free to send diffs to the mailing list for review.

-- 
Emil Sit / MIT CSAIL PDOS / http://pdos.csail.mit.edu/chord/