From owner-stp@oss.sgi.com Sat Oct 14 18:49:01 2000 Received: by oss.sgi.com id ; Sat, 14 Oct 2000 18:48:52 -0700 Received: from pneumatic-tube.sgi.com ([204.94.214.22]:14902 "EHLO pneumatic-tube.sgi.com") by oss.sgi.com with ESMTP id ; Sat, 14 Oct 2000 18:48:45 -0700 Received: from steel.nova.sgi.com (steel.nova.sgi.com [169.238.28.22]) by pneumatic-tube.sgi.com (980327.SGI.8.8.8-aspam/980310.SGI-aspam) via ESMTP id SAA00367 for ; Sat, 14 Oct 2000 18:55:58 -0700 (PDT) mail_from (voellm@sgi.com) Received: from freedom.nova.sgi.com by steel.nova.sgi.com via ESMTP (980427.SGI.8.8.8/930416.SGI) for id VAA94515; Sat, 14 Oct 2000 21:47:27 -0400 (EDT) Date: Sat, 14 Oct 2000 21:45:16 -0400 From: "Anthony F. Voellm" X-Sender: voellm@freedom.nova.sgi.com To: stp@oss.sgi.com Subject: Welcome Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-stp@oss.sgi.com Precedence: bulk Return-Path: X-Orcpt: rfc822;stp-outgoing Hello STP'ers, I just wanted to send a note saying welcome and that I am now the list moderator and point-of-contact at SGI. I have been with SGI for 3.5+ years and have been working on STP/GSN for two years. I have received a couple of source code updates and plan to get these integrated and available soon. Thanks for your participation and feedback. Take care, Tony Anthony F. Voellm Technical Lead - High Performance Network Engineering 14160 Newbrook Dr., Suite 100 Chantilly, VA 20151 (703) 227-8527 Fax (703) 227-8500 http://reality.sgi.com/voellm_nova/ From owner-stp@oss.sgi.com Wed Oct 25 18:14:25 2000 Received: by oss.sgi.com id ; Wed, 25 Oct 2000 18:14:15 -0700 Received: from 255.255.255.255.in-addr.de ([212.8.197.242]:8452 "HELO 255.255.255.255.in-addr.de") by oss.sgi.com with SMTP id ; Wed, 25 Oct 2000 18:14:05 -0700 Received: (qmail 31774 invoked from network); 26 Oct 2000 01:14:02 -0000 Received: from localhost.teuto.de (HELO hermes.marowsky-bree.de) (127.0.0.1) by localhost.teuto.de with SMTP; 26 Oct 2000 01:14:02 -0000 Received: by hermes.marowsky-bree.de (Postfix, from userid 500) id 95179ADB95; Thu, 26 Oct 2000 03:14:02 +0200 (CEST) Date: Thu, 26 Oct 2000 03:14:02 +0200 From: Lars Marowsky-Bree To: stp@oss.sgi.com Subject: Current status? Message-ID: <20001026031402.J8597@marowsky-bree.de> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline Content-Transfer-Encoding: 8bit User-Agent: Mutt/1.2.3i X-Ctuhulu: HASTUR Sender: owner-stp@oss.sgi.com Precedence: bulk Return-Path: X-Orcpt: rfc822;stp-outgoing Good morning everyone, what is the current status of STP for Linux? STP may be interesting to us for network attached storage and replication over the network. Is STP available from inside the kernel? Does it work with 2.2 or 2.4? Please Cc me on the reply, as I am not subscribed to the list. Sincerely, Lars Marowsky-Brée Development HA -- Perfection is our goal, excellence will be tolerated. -- J. Yahl From owner-stp@oss.sgi.com Thu Oct 26 01:23:17 2000 Received: by oss.sgi.com id ; Thu, 26 Oct 2000 01:23:07 -0700 Received: from smtp1.cern.ch ([137.138.128.38]:4103 "EHLO smtp1.cern.ch") by oss.sgi.com with ESMTP id ; Thu, 26 Oct 2000 01:22:38 -0700 Received: from lxplus014.cern.ch (IDENT:root@lxplus014.cern.ch [137.138.161.113]) by smtp1.cern.ch (8.9.3/8.9.3) with ESMTP id KAA04455; Thu, 26 Oct 2000 10:22:31 +0200 (MET DST) Received: from localhost (ppieta@localhost) by lxplus014.cern.ch (8.9.3/8.9.3) with SMTP id KAA27052; Thu, 26 Oct 2000 10:22:30 +0200 X-Authentication-Warning: lxplus014.cern.ch: ppieta owned process doing -bs Date: Thu, 26 Oct 2000 10:22:30 +0200 (CEST) From: Pekka Pietikainen X-Sender: ppieta@lxplus014.cern.ch To: Lars Marowsky-Bree cc: stp@oss.sgi.com Subject: Re: Current status? In-Reply-To: <20001026031402.J8597@marowsky-bree.de> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-stp@oss.sgi.com Precedence: bulk Return-Path: X-Orcpt: rfc822;stp-outgoing On Thu, 26 Oct 2000, Lars Marowsky-Bree wrote: > Good morning everyone, > > what is the current status of STP for Linux? Hi 102MB/s with 1MB AceNIC's with jumbo frames using < 5% CPU is what I currently get ;) > STP may be interesting to us for network attached storage and replication over > the network. > > Is STP available from inside the kernel? Does it work with 2.2 or 2.4? The current code assumes that the protocol is used from userland (it tries to use map_user_kiobuf for every case). I'm looking at fixing that soon, though using nbd as a testcase. From owner-stp@oss.sgi.com Thu Oct 26 09:03:19 2000 Received: by oss.sgi.com id ; Thu, 26 Oct 2000 09:03:09 -0700 Received: from laime.cs.uchicago.edu ([128.135.11.244]:31483 "EHLO laime.cs.uchicago.edu") by oss.sgi.com with ESMTP id ; Thu, 26 Oct 2000 09:02:53 -0700 Received: from candide.cs.uchicago.edu (candide.cs.uchicago.edu [128.135.11.62]) by laime.cs.uchicago.edu (8.10.2/8.9.3) with SMTP id e9QG2lU15389 for ; Thu, 26 Oct 2000 11:02:47 -0500 (CDT) Received: by candide.cs.uchicago.edu (5.57/4.7) id AA24147; Thu, 26 Oct 00 11:01:01 -0500 Message-Id: <10010261601.AA24147@candide.cs.uchicago.edu> To: stp@oss.sgi.com Subject: Re: Current status? In-Reply-To: Message from Pekka Pietikainen of "Thu, 26 Oct 2000 10:22:30 +0200." Date: Thu, 26 Oct 2000 11:02:07 -0500 From: Stephen Bailey Sender: owner-stp@oss.sgi.com Precedence: bulk Return-Path: X-Orcpt: rfc822;stp-outgoing Hi Pekka (et al), > 102MB/s with 1MB AceNIC's with jumbo frames using < 5% CPU is what I > currently get ;) I believe we have to take jumbo frames off the table for STP. As I understand it (somebody please correct me if I'm wrong!), 802 has killed jumbo frames twice, in recent history, once for 1 GigE and once for 10 GigE. That basically means that even though some vendors support it (Alteon and Extreme, that I know of), it's not going to happen. I have heard this on from two different sources separated by years, most recently a few weeks ago from somebody at 3Com who should know. He did say that he thought that Alteon was trying to get an IETF standardization of jumbo frames, but that seems (to him and me) unlikely to go anywhere, because it's not the right place for it to be decided. I guess you could say that running jumbo ethernet is like running Myranet or something---an isolated, nonstandard networking technology. In other words, SURE, some people do it, but it's considered fringe rather than commodity. I think the upshot is that if STPers focus on jumbo frame performance, we're going to look like we're completely out of touch with the real world. So (I'm almost afraid to ask) what's the performance for 1500 byte MTU (1K STUs)? The good news is that, you'll probably be able to say < 2% CPU, and the bad news is that the data rate is going to be, well, lower too. Steph From owner-stp@oss.sgi.com Thu Oct 26 09:46:09 2000 Received: by oss.sgi.com id ; Thu, 26 Oct 2000 09:45:59 -0700 Received: from sheffield.concentric.net ([207.155.252.12]:43229 "EHLO sheffield.cnchost.com") by oss.sgi.com with ESMTP id ; Thu, 26 Oct 2000 09:45:50 -0700 Received: from atheros.com (w231.z064001104.sjc-ca.dsl.cnc.net [64.1.104.231]) by sheffield.cnchost.com id MAA02293; Thu, 26 Oct 2000 12:45:48 -0400 (EDT) [ConcentricHost SMTP Relay 1.10] Message-ID: <39F86063.5CF33E6D@atheros.com> Date: Thu, 26 Oct 2000 09:48:35 -0700 From: Aman Singla Organization: Atheros Communications, Inc. X-Mailer: Mozilla 4.75 [en] (X11; U; Linux 2.2.14-5.0 i686) X-Accept-Language: en MIME-Version: 1.0 To: Stephen Bailey CC: stp@oss.sgi.com Subject: Re: Current status? References: <10010261601.AA24147@candide.cs.uchicago.edu> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-stp@oss.sgi.com Precedence: bulk Return-Path: X-Orcpt: rfc822;stp-outgoing > I believe we have to take jumbo frames off the table for STP. As I > understand it (somebody please correct me if I'm wrong!), 802 has > killed jumbo frames twice, in recent history, once for 1 GigE and once > for 10 GigE. That basically means that even though some vendors > support it (Alteon and Extreme, that I know of), it's not going to > happen. > > I have heard this on from two different sources separated by years, > most recently a few weeks ago from somebody at 3Com who should know. > He did say that he thought that Alteon was trying to get an IETF > standardization of jumbo frames, but that seems (to him and me) > unlikely to go anywhere, because it's not the right place for it to be > decided. > > I guess you could say that running jumbo ethernet is like running > Myranet or something---an isolated, nonstandard networking > technology. In other words, SURE, some people do it, but it's > considered fringe rather than commodity. > > I think the upshot is that if STPers focus on jumbo frame performance, > we're going to look like we're completely out of touch with the real > world. > > So (I'm almost afraid to ask) what's the performance for 1500 byte MTU > (1K STUs)? The good news is that, you'll probably be able to say < 2% > CPU, and the bad news is that the data rate is going to be, well, > lower too. I totally agree with you on the jumbo frames thing. I remember doing an extensive analysis of the performance using 1K STUs on alteon's acenic. I was getting close to 85MB/s - and the bottleneck was the firmware loop times in the card. There's been some user level I/O work at University of Cambridge(Ian Pratt) - which is very much related to STP. They rewrote firmware for the acenic card, and report close to wire speed with standard ethernet sized frames. Their approach is very similar to STP, and I'd venture that same results would hold for STP - if someone rewrote the firmware! :a From owner-stp@oss.sgi.com Thu Oct 26 14:43:44 2000 Received: by oss.sgi.com id ; Thu, 26 Oct 2000 14:43:34 -0700 Received: from 255.255.255.255.in-addr.de ([212.8.197.242]:18189 "HELO 255.255.255.255.in-addr.de") by oss.sgi.com with SMTP id ; Thu, 26 Oct 2000 14:43:04 -0700 Received: (qmail 14768 invoked from network); 26 Oct 2000 21:43:02 -0000 Received: from localhost.teuto.de (HELO hermes.marowsky-bree.de) (127.0.0.1) by localhost.teuto.de with SMTP; 26 Oct 2000 21:43:02 -0000 Received: by hermes.marowsky-bree.de (Postfix, from userid 500) id 01EC4AD667; Thu, 26 Oct 2000 23:42:59 +0200 (CEST) Date: Thu, 26 Oct 2000 23:42:59 +0200 From: Lars Marowsky-Bree To: Pekka Pietikainen Cc: stp@oss.sgi.com Subject: Re: Current status? Message-ID: <20001026234259.F14490@marowsky-bree.de> References: <20001026031402.J8597@marowsky-bree.de> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline Content-Transfer-Encoding: 8bit User-Agent: Mutt/1.2.3i In-Reply-To: ; from "Pekka Pietikainen" on 2000-10-26T10:22:30 X-Ctuhulu: HASTUR Sender: owner-stp@oss.sgi.com Precedence: bulk Return-Path: X-Orcpt: rfc822;stp-outgoing On 2000-10-26T10:22:30, Pekka Pietikainen said: > > what is the current status of STP for Linux? > Hi > 102MB/s with 1MB AceNIC's with jumbo frames using < 5% CPU is what I > currently get ;) Sounds great. How does the average latency compare to TCP ? > > Is STP available from inside the kernel? Does it work with 2.2 or 2.4? > The current code assumes that the protocol is used from userland > (it tries to use map_user_kiobuf for every case). I'm looking at fixing > that soon, though using nbd as a testcase. Exactly, nbd (or rather drbd ;), was what I was thinking about using STP for. It isn't that urgent, but good to see someone is working on it ;-) Sincerely, Lars Marowsky-Brée Development HA -- Perfection is our goal, excellence will be tolerated. -- J. Yahl From owner-stp@oss.sgi.com Thu Oct 26 15:13:24 2000 Received: by oss.sgi.com id ; Thu, 26 Oct 2000 15:13:03 -0700 Received: from ccsalpha3.nrl.navy.mil ([132.250.112.51]:517 "EHLO ccsalpha3.nrl.navy.mil") by oss.sgi.com with ESMTP id ; Thu, 26 Oct 2000 15:12:45 -0700 Received: from ncst.nrl.navy.mil (master.nrl.navy.mil [128.60.24.130]) by ccsalpha3.nrl.navy.mil (8.9.3/8.9.3) with ESMTP id SAA31704 for ; Thu, 26 Oct 2000 18:12:42 -0400 (EDT) Message-ID: <39F86412.7D5A94F8@ncst.nrl.navy.mil> Date: Thu, 26 Oct 2000 17:04:18 +0000 From: Eric Sydow Organization: Naval Research Lab X-Mailer: Mozilla 4.75 [en] (X11; U; Linux 2.4.0-test7 i686) X-Accept-Language: en MIME-Version: 1.0 To: stp@oss.sgi.com Subject: STP over ATM??? Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-stp@oss.sgi.com Precedence: bulk Return-Path: X-Orcpt: rfc822;stp-outgoing Hello, Has anyone ever thought about expanding the range of hardware STP operates on? Specifically ATM NICs? Specifically the Fore HE cards? I like the idea of STP but I need it to work over a long haul & ethernet obviously doesn't travel well (esp. GigE). Just throwing it out there :-) Eric From owner-stp@oss.sgi.com Thu Oct 26 15:26:54 2000 Received: by oss.sgi.com id ; Thu, 26 Oct 2000 15:26:34 -0700 Received: from genroco.com ([205.254.195.202]:31506 "EHLO gifw.genroco.com") by oss.sgi.com with ESMTP id ; Thu, 26 Oct 2000 15:26:32 -0700 Received: from gi2.genroco.com (gi2.genroco.com [192.133.120.3]) by gifw.genroco.com (8.9.3/8.9.3) with ESMTP id RAA29057; Thu, 26 Oct 2000 17:26:30 -0500 Received: from don (don.genroco.com [192.133.120.101]) by gi2.genroco.com (8.9.3/8.9.3) with SMTP id RAA23154; Thu, 26 Oct 2000 17:26:30 -0500 From: "Don Woelz" To: "Eric Sydow" , Subject: RE: STP over ATM??? Date: Thu, 26 Oct 2000 17:26:30 -0500 Message-ID: MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook IMO, Build 9.0.2416 (9.0.2911.0) Importance: Normal In-Reply-To: <39F86412.7D5A94F8@ncst.nrl.navy.mil> X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 Sender: owner-stp@oss.sgi.com Precedence: bulk Return-Path: X-Orcpt: rfc822;stp-outgoing Hi all, FYI, GENROCO is writing STP stacks for Compaq Tru64 UNIX, Sun Solaris, IBM AIX, and will soon be doing Linux, NT, and HP/UX. These are intended to run with our PCI GSN NIC as well as the standard GigE NICs supported by these platforms. We are doing this primarily to support the SCSI on STP (SST) protocol to enable Gigabit Ethernet SANs, but that should not be a limitation to using ST for other purposes. We bridge TCP/IP and STP between GigE, OC48 POS (a long haul option), HIPPI, GSN, and (for SST) Fibre Channel in our hardware products. We'll be showing a lot of this in our booth (#856) at SC2000 in Dallas (Nov. 6 - 9). If you are refering to STP acceleration in the NICs, our PCI GSN NIC will support STP acceleration soon. Don Donald D. Woelz Tel: 262-644-2505 V. P. of Marketing Mobile: 414-732-7705 GENROCO, Inc. Toll Free: 800-243-6762 255 Info Hwy Fax: 262-644-6667 Slinger, WI 53086 > -----Original Message----- > From: owner-stp@oss.sgi.com [mailto:owner-stp@oss.sgi.com]On Behalf Of > Eric Sydow > Sent: Thursday, October 26, 2000 12:04 PM > To: stp@oss.sgi.com > Subject: STP over ATM??? > > > Hello, > > Has anyone ever thought about expanding the range of hardware STP > operates on? Specifically ATM NICs? Specifically the Fore HE cards? > > I like the idea of STP but I need it to work over a long haul & ethernet > obviously doesn't travel well (esp. GigE). > > Just throwing it out there :-) > > Eric > > From owner-stp@oss.sgi.com Fri Oct 27 00:53:39 2000 Received: by oss.sgi.com id ; Fri, 27 Oct 2000 00:53:29 -0700 Received: from smtp1.cern.ch ([137.138.128.38]:15628 "EHLO smtp1.cern.ch") by oss.sgi.com with ESMTP id ; Fri, 27 Oct 2000 00:53:10 -0700 Received: from cern.ch (IDENT:ben@lxplus009.cern.ch [137.138.161.118]) by smtp1.cern.ch (8.9.3/8.9.3) with ESMTP id JAA01615; Fri, 27 Oct 2000 09:53:00 +0200 (MET DST) X-Authentication-Warning: smtp1.cern.ch: Host IDENT:ben@lxplus009.cern.ch [137.138.161.118] claimed to be cern.ch Message-ID: <39F9345B.F037F2D2@cern.ch> Date: Fri, 27 Oct 2000 09:52:59 +0200 From: Ben Segal Organization: CERN X-Sender: "Ben Segal" <@smtp.cern.ch> (Unverified) X-Mailer: Mozilla 4.75C-CERN UNIX lxplus009 45 [en] (X11; U; Linux 2.2.12-20smp i686) X-Accept-Language: fr-FR, en MIME-Version: 1.0 To: Eric Sydow CC: stp@oss.sgi.com, Arie Van Praag Subject: Re: STP over ATM??? References: <39F86412.7D5A94F8@ncst.nrl.navy.mil> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-stp@oss.sgi.com Precedence: bulk Return-Path: X-Orcpt: rfc822;stp-outgoing Eric Sydow wrote: > > Hello, > > Has anyone ever thought about expanding the range of hardware STP > operates on? Specifically ATM NICs? Specifically the Fore HE cards? > > I like the idea of STP but I need it to work over a long haul & ethernet > obviously doesn't travel well (esp. GigE). > > Just throwing it out there :-) > > Eric Hi Eric, Actually GigE _is_ going to travel well, it's already begun in fact. DWDM links are going to be the mainstream long haul fibre technology, with SDH declining, and GigE/10GigE travels perfectly on that. Otherwise see Don Woelz's comments on STP over OC-48c, etc. By the way, Don forgot to mention that CERN works closely with Genroco on their bridge product, in fact we built the GigE and OC-48c bridge blades as part of a joint development project. Ben -- Ben M. Segal / CERN - IT Division, 1211 Geneva 23, Switzerland. Tel: +41-22 767 4941 / Mobile: +41-79 201 0618 Fax: +41-22 767 7155 WWW: http://home.cern.ch/ben/ From owner-stp@oss.sgi.com Fri Oct 27 06:24:50 2000 Received: by oss.sgi.com id ; Fri, 27 Oct 2000 06:24:40 -0700 Received: from smtp1.cern.ch ([137.138.128.38]:25092 "EHLO smtp1.cern.ch") by oss.sgi.com with ESMTP id ; Fri, 27 Oct 2000 06:24:25 -0700 Received: from lxplus003.cern.ch (IDENT:root@lxplus003.cern.ch [137.138.161.124]) by smtp1.cern.ch (8.9.3/8.9.3) with ESMTP id PAA31511 for ; Fri, 27 Oct 2000 15:24:18 +0200 (MET DST) Received: from localhost (ppieta@localhost) by lxplus003.cern.ch (8.9.3/8.9.3) with SMTP id PAA06364 for ; Fri, 27 Oct 2000 15:24:17 +0200 X-Authentication-Warning: lxplus003.cern.ch: ppieta owned process doing -bs Date: Fri, 27 Oct 2000 15:24:17 +0200 (CEST) From: Pekka Pietikainen X-Sender: ppieta@lxplus003.cern.ch Reply-To: Pekka Pietikainen To: stp@oss.sgi.com Subject: Re: Current status? In-Reply-To: <10010261601.AA24147@candide.cs.uchicago.edu> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-stp@oss.sgi.com Precedence: bulk Return-Path: X-Orcpt: rfc822;stp-outgoing On Thu, 26 Oct 2000, Stephen Bailey wrote: > I think the upshot is that if STPers focus on jumbo frame performance, > we're going to look like we're completely out of touch with the real > world. > > So (I'm almost afraid to ask) what's the performance for 1500 byte MTU > (1K STUs)? The good news is that, you'll probably be able to say < 2% > CPU, and the bad news is that the data rate is going to be, well, > lower too. ./gen4 starting on toy3 ( reclen=524288 TCPhost=192.168.9.2 ) - Fri Oct 27 14:16:18 2000 # description host sample_KB total_MB sample_KB/s avge_KB/s cpu_sec user_sec sys_sec sec/MB cpu_pct 1 source toy3 209715.203 209.715 47240.516 47240.516 0.060 0.010 0.050 0.000 1 1 source toy3 209715.203 419.430 47267.035 47253.775 0.090 0.000 0.090 0.000 2 1 source toy3 209715.203 629.146 47297.066 47268.206 0.040 0.000 0.040 0.000 1 2k STU's give around 80MB and 4k is over 100. The hardware used is a dual pIII/500 with 66MHz PCI sending to a pII/400 with standard frames, the other way around the performance is slightly lower. To give a comparison, TCP gets about 62M/s with ~= 100% CPU on my hardware. As for latency, [root@toy3 i686-linux]# ./lat_stp 192.168.9.2 STP latency using 192.168.9.2: 569.2200 microseconds [root@toy3 i686-linux]# ./lat_tcp 192.168.9.2 TCP latency using 192.168.9.2: 149.6932 microseconds Which isn't too bad considering STP needs that extra round-trip for every write() to setup the buffers. From owner-stp@oss.sgi.com Fri Oct 27 08:01:21 2000 Received: by oss.sgi.com id ; Fri, 27 Oct 2000 08:01:11 -0700 Received: from laime.cs.uchicago.edu ([128.135.11.244]:12488 "EHLO laime.cs.uchicago.edu") by oss.sgi.com with ESMTP id ; Fri, 27 Oct 2000 08:00:58 -0700 Received: from candide.cs.uchicago.edu (candide.cs.uchicago.edu [128.135.11.62]) by laime.cs.uchicago.edu (8.10.2/8.9.3) with SMTP id e9RF0vU18359 for ; Fri, 27 Oct 2000 10:00:57 -0500 (CDT) Received: by candide.cs.uchicago.edu (5.57/4.7) id AA25349; Fri, 27 Oct 00 09:59:11 -0500 Message-Id: <10010271459.AA25349@candide.cs.uchicago.edu> To: stp@oss.sgi.com Subject: Re: Current status? In-Reply-To: Message from Pekka Pietikainen of "Fri, 27 Oct 2000 15:24:17 +0200." Date: Fri, 27 Oct 2000 10:00:14 -0500 From: Stephen Bailey Sender: owner-stp@oss.sgi.com Precedence: bulk Return-Path: X-Orcpt: rfc822;stp-outgoing Pekka, Thanks for the numbers. 80 MB/s with 1-2% CPU is pretty cool. As Aman mentioned, you've hit the limit of that poor little embedded processor on the NIC, at least with your present firmware (heh, heh, heh, thin adapters rule!). > As for latency, > [...] > Which isn't too bad considering STP needs that extra round-trip for > every write() to setup the buffers. Just for those who don't know, the latency comparison would be more appropriately made using an ST persistent memory regions, to avoid the setup round-trip, and OS context switches. ST persistent memory is the feature in ST to which the marketing label `low latency' applies. I don't know if the Linux STP code supports persistent memory yet, but perhaps not. Steph From owner-stp@oss.sgi.com Tue Oct 31 08:02:30 2000 Received: by oss.sgi.com id ; Tue, 31 Oct 2000 08:02:20 -0800 Received: from blake.servalan.org ([195.40.110.2]:38662 "EHLO blake.servalan.org") by oss.sgi.com with ESMTP id ; Tue, 31 Oct 2000 08:02:09 -0800 Received: from jenna (testy.servalan.org [195.40.110.49]) by blake.servalan.org (8.11.0/8.11.0) with SMTP id e9VG1xZ06830 for ; Tue, 31 Oct 2000 16:01:59 GMT Message-ID: <00bb01c04354$00fd7ba0$316e28c3@servalan.org> From: "John Thorp" To: Subject: Network perforance and low CPU usage Date: Tue, 31 Oct 2000 16:02:43 -0000 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.50.4133.2400 X-MIMEOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 Sender: owner-stp@oss.sgi.com Precedence: bulk Return-Path: X-Orcpt: rfc822;stp-outgoing I have just started looking at a project which needs to process high data volumes. The idea is to string together a set of machines connected by a Gigabit network. A typical data rate would be a sustained 350Mbits/s input and output for each machine. The machines we are going to use will be 4 CPU 700Mhz running Linux. At the moment I am trying to do this with as little CPU being used for the network I/O as possible. It looks like STP would be ideal for this application. We are looking at using the Netgear GA620 network card - is this the best? Has anyone already done this (I/O both on and off the machine at the same time) who is willing to give a potential CPU loading? John Thorp