MORE: Experimenting with #sendUDPData:toHost:port:

Bob Arning arning at charm.net
Wed Dec 6 22:52:33 UTC 2000


On Wed, 06 Dec 2000 17:42:10 -0500 John Chludzinski <jchludzinski at worldkey.net> wrote:
>I just tried using #sendData:toHost:port: in lieu of #sendUDPData:toHost:port: in the #clientServerTestUDP example in the Socket class. 
>(I'm STILL experimenting) .  This time I got:
>
>client/server UDP test done; time = 6560
>2500 packets, 10000000 bytes sent (1524390 Bytes/sec)
>2500 packets, 10000000 bytes received (1524390 Bytes/sec)
>4000 bytes/packet, 381 packets/sec, 0 packets dropped
>
>Using the #sendData: with the socket attached to a specific IP/port pair (in the #clientServerTestUDP example), I got:
>
>client/server UDP test done; time = 221
>2500 packets, 10000000 bytes sent (45248868 Bytes/sec)
>2500 packets, 10000000 bytes received (45248868 Bytes/sec)
>4000 bytes/packet, 11312 packets/sec, 0 packets dropped
>
>The thing that obviously stands out is the difference in times: 6560 vs. 221.
>
>Any ideas why?

Sure, in the first example you set the host and port for every packet which filters down to 2500 connect requests. In the second case you connect once and reuse the connection 2500 times.

Cheers,
Bob





More information about the Squeak-dev mailing list