tcp vs. udp queries performance difference

[phn at icke-reklam.ipsec.nu]

KyoungSoo Park <kyoungso at cs.princeton.edu> wrote:
> Hi,

> I'm not sure if this is relevent to this list, but I have a simple=20
> question.
> Is it generally better (in terms of overall latency) to use persistent=20
> tcp queries to a nameserver (over typical udp queries)
> when you have a high volume of names to lookup consistently ? ( I mean=20
> A-record lookup from a normal client.)
> For example, when you're running a proxy server, you get a higher=20
> pressure of name lookups, so
> I think it's better to use a persistent tcp connection to a nameserver=20
> rather than getting screwed by udp packet losses.

> However, I tested this, but I don't see big difference between two meth=
ods.
> Even worse, the persistent connection's overall latency seems worse tha=
n=20
> the udp queries in many cases.
> Is there anyone who's already experienced this and possibly knows why?

You cannot do better in terms of latency then UDP. The drawback
might be lack of flow-control, but it seems unlikley that=20
that region of opertion is made.

It's also less CPU involved when using UDP.

> Thanks,
> KyougnSoo

> PS. I sent this to bind9-users too,  Sorry if you see this twice.



--=20
Peter H=E5kanson        =20
        IPSec  Sverige      ( At Gothenburg Riverside )
           Sorry about my e-mail address, but i'm trying to keep spam out=
,
	   remove "icke-reklam" if you feel for mailing me. Thanx.