From: Marko Mäkelä (marko.makela_at_hut.fi)
Date: 2004-03-10 08:39:55
On Tue, Mar 09, 2004 at 11:35:12PM +0000, Davison, Lee wrote: > The idea was to simulate, as closely as I can, many unsynchronised > requests from many places. Flood ping couldn't send two or more > simultaneous (or as near as makes no odds) requests and so would > never fill the Vic's receive buffer. True, I didn't think of that. But are the requests unsynchronised enough? By default, ping sends ICMP echo request at roughly one-second intervals. You would have to run it very long in order to get enough clock drift between the ping clients, to make the requests arrive close to each other. BTW, couldn't you get a better stress test by modifying the TCP/IP stack on the sender, to send multiple ICMP echo request packets in a row? Flood ping would measure a different thing: throughput. I'd be interested in the results. I've noticed that Windows machines have hard time keeping up with flood ping, while Linux loses practically no packets. That is on a 100 Mb/s full duplex link, on a several hundred MHz processor. It'd be interesting to see how much the Vic-20 can manage using no DMA. Marko Message was sent through the cbm-hackers mailing list
Archive generated by hypermail pre-2.1.8.