[Spread-users] Configuration file answer

Aswin Almeida aalmeida at bbn.com
Fri Apr 4 12:41:37 EST 2003

Hi Yair,

Comments are inline.

>To me, the segment definition in your example below seems incorrect.
>You should not include in a spread configuration file unless
>the configuration file includes exactly one segment with exactly one

Right, this is what I was thinking as well.

While helping Brian Beck recently (recall he was the Red team fellow asking 
us questions about the integrated architecture), I noticed that he had specified in his spread.conf file.

This made me think about our own experiment runs.  Each run's conf files 
specify as part of each segment.

I was concerned that this would cause side effects that you, folks at JHU, 
or Jonathan would be aware of.  So, airing on the side of caution, I posed 
the question.  If it's simply not supposed to work, well, we've done many, 
many runs with as part of one or more segments that are 
a.b.c.255, x.y.z.255, etc which contain 2 or more machines.

>Having said that, I don't think this relates in any way to the problem
>you have with partitioning.


I threw that in at the end because I'd knew I'd have your attention and 
wanted you to see that Sara and I had discovered quite a bit of info beyond 
what was discussed at the TIC <:  Didn't mean that the two things were 
necessarily tied together.



>On Friday, April 04, 2003 11:35 AM
>Aswin Almeida aalmeida at bbn.com wrote:
>Aswin> Yair,
>Aswin> The reason I ask is because we have this working in lab tests and 
>Aswin> experiment runs.
>Aswin> In fact, our experiment tool dynamically creates the spread.conf 
>files and
>Aswin> segments on a per enclave basis (only including machines that are
>Aswin> involved).  I was curious if this might cause any side effects.
>Aswin> Example:
>Aswin> Spread_Segment {
>Aswin>          localhost
>Aswin>          cam100
>Aswin>          cam101
>Aswin> }
>Aswin> Spread_Segment {
>Aswin>          localhost
>Aswin>          afrl3
>Aswin>          afrl4
>Aswin> }
>Aswin> The partitioning issue that we are seeing is *not* confined to just 
>the TIC
>Aswin> and its outbound links.  It occurs with other sites, even when the 
>Aswin> loss rate from "s" and "r" is 5-7%.  This suggests that the problem is
>Aswin> either with the network itself (a possibility, as one does not 
>control the
>Aswin> public links we are relying on for the VPN) or we'd still be able 
>to tweak
>Aswin> something within Spread (the application itself).
>Aswin> See Sara's reply (previously) for details.
>Aswin> Regards,
>Aswin> Aswin
>Aswin> At 07:45 AM 04/03/2003 -0500, Yair Amir wrote:
> >>Hi,
> >>
> >>This configuration should not work. Take out the from the
> >>segment.
> >>
> >>         Cheers,
> >>
> >>         :) Yair.
> >>
> >> > Folks,
> >> >
> >> > One follow up question to this:
> >> >
> >> > Is there anything wrong with specifying a localhost entry for a Spread
> >> > Segment that defines a.b.c.255?
> >> >
> >> > For example:
> >> >
> >> > Spread_Segment x.2.2.255 {
> >> >         localhost     <---
> >> >         other1     
> >> >         other2     
> >> > }
> >> >
> >> > -Aswin
> >>
> >>
> >>_______________________________________________
> >>Spread-users mailing list
> >>Spread-users at lists.spread.org
> >>http://lists.spread.org/mailman/listinfo/spread-users
>Aswin> _______________________________________________
>Aswin> Spread-users mailing list
>Aswin> Spread-users at lists.spread.org
>Aswin> http://lists.spread.org/mailman/listinfo/spread-users
>Spread-users mailing list
>Spread-users at lists.spread.org

More information about the Spread-users mailing list