(Illustration by Gaich Muramatsu)
On Thu, 19 Feb 2004, Brett Lymn wrote: > On Tue, Feb 17, 2004 at 10:54:54PM -0500, M. Satyanarayanan wrote: > > The right thing to aim for is a tasteful job of fixing RPC2. > If there was to be a version bump for RPC2 then perhaps it would be an > ideal time to address some LP64 issues as well? I am not certain if One more thing about RPC2, it looks like sftp can't cope with "long fat networks" which gives poor performance as soon as there is remarkable latency to the server. (my observations both on overseas connections and with dsl at home, both yielding similar results) The bandwidth of the home dsl connection is about 700 Kbyte/s but at 100 ms latency venus fetched files at 40-50 K/s, then when the provider changed the routing and the latency reduced to 45 ms, I get usually ~135 K/s, or, occasionally, ~300 K/s. This data is not at all exact, but the numbers are in that order... For comparison, when a client and the server are connected to the same 100 Mbit switch, the speed at fetching is 6.5...10 Mbyte/s. This problem is not a "vital" thing, but might be worth to address if the protocol is going to change anyway... For a global file system performance on long fat lines can be important, even with Coda's efficient caching. My 2c, -- IvanReceived on 2004-02-19 08:04:37