Re: replacing ICMP pinging with SONAR

map@iphil.net writes:

>Hi,
>
>Duane Wessels wrote:
>> Since I originally looked at SONAR I've found a couple reasons to
>> NOT replace the 'pinger' with SONAR. One is that Squid would lose
>> the ability to send source pings at ICMP packets.
>
>Wouldn't it be the same if the SONAR server were running on the
>same machine as Squid, then Squid just sends the query?

Not quite sure what you mean, but I think you mean building the
SONAR server into Squid? That would be taking what we have now
and adding an interface whereby the ICMP data could be queried with
the SONAR protocol.

One of my reasons for looking at SONAR was so that there would
be less code in Squid, not more :-)

>> The other reason is that SONAR only defines the message format between
>> a SONAR server and client. It does not (yet) define how the metric is
>> calculated.
>
>Yeah it's still a work in progress, it seems. Looks like the simplest
>metric is the RTT.
>
>> So, in general, it is not possible to compare two SONAR metrics
>> from two SONAR servers (for the same destination).
>
>This is what I was hoping for - some way to triangulate the "location"
>of a server. If you can simplify the Internet into "local" and
>"foreign" components, the way I can, since the two are separated by congested
>links, it would be easier. I compare the SONAR result from a local
>server to a result from the US, and decide whether I should use a local
>or a foreign parent.

I think for the next stage of SONAR implementation, they will add the
ability to request the response in "milliseconds of RTT" or whatever so
that we actually could compare results from different sources.

Duane W.
Received on Tue Jul 29 2003 - 13:15:40 MDT