14 messages in com.perforce.perforce-userp4Win with a UNIX server| From | Sent On | Attachments |
|---|---|---|
| Anth...@pilotsw.com | 25 Aug 1998 12:09 | |
| Joha...@xpress.se | 25 Aug 1998 12:34 | |
| Babu...@Mitel.COMBabu_Somisetty | 25 Aug 1998 12:59 | |
| Anth...@pilotsw.com | 25 Aug 1998 13:41 | |
| WesP...@softweyr.com | 25 Aug 1998 21:03 | |
| WesP...@softweyr.com | 25 Aug 1998 22:15 | |
| Nick...@pobox.com | 26 Aug 1998 04:52 | |
| Anth...@pilotsw.com | 26 Aug 1998 06:21 | |
| Fred...@mydata.se | 26 Aug 1998 06:34 | |
| Raym...@orion.no | 26 Aug 1998 06:55 | |
| RobC...@within.com | 26 Aug 1998 07:25 | |
| Rich...@geodesic.com | 26 Aug 1998 08:08 | |
| Nick...@pobox.com | 26 Aug 1998 08:21 | |
| Anth...@pilotsw.com | 27 Aug 1998 05:37 |
| Subject: | p4Win with a UNIX server![]() |
|---|---|
| From: | Nick...@pobox.com (Nick...@pobox.com) |
| Date: | 08/26/1998 04:52:01 AM |
| List: | com.perforce.perforce-user |
p4win seems to do a large number of p4 transactions when starting up. At first I assumed it was just doing "p4 files //...", or similar, to build its directory hierarchy, but the speed doesn't scale linearly with the connection bandwidth, so my guess is that it does a large number of transactions, possibly even one for each file. That way the scaling is by transaction latency (setting up and tearing down a p4 transaction), not bandwidth (just listening to a stream), which fits our observations.
With a fair-sized depot (p4 files //... | wc says 116296 files), served from FreeBSD, across a slowish transatlantic connection (IP latency 300-1000ms, bandwidth 10-30KB/sec), p4win can take 30-45 _minutes_ to start up. On a machine on the same ethernet as the server it is much faster but still several minutes. We have reported this to Perforce. We haven't tried with an NT server (why would anyone want to serve anything from NT? :->)
Nick Barnes




