[ldns-users] Question about ldns packet timings

[Ryan Braud]

Ideally, yes, you would use kernel timestamps for both sent and received
packets.  However, ldns does not give you (me) the ability to do this from
queries sent via ldns_resolver_query() since it keeps the sockets
internally.  In theory, I could just build the packet using ldns and send
the packets myself on sockets I create (so that I can set the proper socket
options), but it would be nice if ldns did this for me.

Ryan

On Fri, Mar 20, 2015 at 2:10 PM, Robert Edmonds <edmonds at debian.org> wrote:

> Ryan Braud wrote:
> > Hi everyone, I'm not sure if this is the correct list for this question,
> > but I had a question/enhancement request for ldns.  Currently, the
> function
> > ldns_pkt_querytime() returns the amount of time a request took, measured
> > with gettimeofday().  I was wondering if there was a good reason why this
> > time is not calculated from a kernel timestamp via the SIOCGSTAMP ioctl?
> > We are using libldns in our production environment and when the processor
> > gets busy, we end up with random spikes in latency due to scheduling
> > overhead, etc.
>
> Hi, Ryan:
>
> Doesn't SIOCGSTAMP only give you the timestamp of the last received
> packet?  If you want the "real" request latency, wouldn't you want
> kernel timestamps on both the sent and received packets?
>
> (I think you can do this, at least on Linux, with SO_TIMESTAMPING.)
>
> --
> Robert Edmonds
> edmonds at debian.org
> _______________________________________________
> ldns-users mailing list
> ldns-users at open.nlnetlabs.nl
> http://open.nlnetlabs.nl/mailman/listinfo/ldns-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.nlnetlabs.nl/pipermail/ldns-users/attachments/20150320/d2a1c3ff/attachment.htm>