Date: Thu, 29 Jan 1998 11:10:40 -0100 (GMT) From: Remy NONNENMACHER <remy@synx.com> To: sthaug@nethelp.no Cc: freebsd-smp@FreeBSD.ORG Subject: Re: seeking for perfs... Message-ID: <Pine.A32.3.91.980129105555.16326T-100000@rs1> In-Reply-To: <13604.886015934@verdi.nethelp.no>
index | next in thread | previous in thread | raw e-mail
On Wed, 28 Jan 1998 sthaug@nethelp.no wrote: > > For now, i would like to now if i can do something to speed-up the > > network. The machine runs DFE500TX 100Mbps FDX card connected to a > > switch. My questions are : > > > > Is the 58Mbps i get a normal rating ? Can i get more ? Using what card ? > > You should be able to come very close to saturating a 100 Mbps Ethernet > with that card, unless you have a card with the 21140-AA chip. Later chip > versions are better (the AA version can't read multiple words in one PCI > bus transaction). > > Using ttcp, that comes out to 100 * 1440/1538 = 93.6 Mbit/s application > to application, or 100 * 1460/1538 = 94.9 Mbit/s if you turn off RFC 1323 > and RFC 1644 support. I've measured 93.57 Mbit/s (with 1323/1644 support > turned on) myself - between a PPro-200 (21140 card) and a P-133 (Pro 100/B > card). > You were right. I just checked this using a flooder and got a 8220 paq/s and somewhere between 95 and 98 Mbit/s. My problem seems to be that using ftp, rcp or nfs, i can go more than 58Mbit/s. I suspect the problem is the giant lock position since, for example, when using a reading process and a writing one (through pipe) i never have two running processes and perfs are half the one i expect. However, let's be happy with that. SMP is a pleasure and an impressive work. Thanks to the team !!. I wonder what will be the result when locking will be of fine granularity. Is the current silent in the SMP list the sign that SNAP-971225 is quiet stable ?home | help
Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.A32.3.91.980129105555.16326T-100000>