- WLUG - wlug.mailman3.com

Re: poor PCMCIA network performance

by Mark J. Dulcey

Robert L Krawitz wrote: > I use two network cards in my Dell Inspiron 8000 (PIII-700): a 3Com > 3CCFE574BT and a Linksys wireless card. With the 3Com card (which is > a 10/100) I never get better than 1.4 MB/sec (about 12 Mb/sec), even > when the card is in 100 Mb mode (at some point in the past the limit > was about 2.5 MB/sec). At that rate, top shows my CPU utilization > pegged at about 100%, even doing a simple ftp. I know the card > doesn't do DMA, but that's rather slow even for a PIO transfer. These > cards have been around for ages; I don't think people would have been > too satisfied with 400 KB/sec saturating a 200 MHz system. Your results with the 3Com card are normal for a 16-bit PCMCIA card. The wall you're hitting is the PCMCIA interface. You would have done much better to get a CardBus card (such as the 3CCFE575CT); you should be able to saturate a Fast Ethernet with one of those, and the CPU utilization would be lower. Basically, 10/100 16-bit cards are something of a lie, since you can never get anywhere near 100 Mbps from them. The same is true of 10/100 USB interfaces (the 12 Mbps speed of USB is a hard limit), though USB 2.0 (480 Mbps) could be different if anyone ever produces a USB 2.0 Ethernet interface. > On the Linksys card, I likewise never get more than about 250 KB/sec, > even if the computer is right next to the WAP and the link quality is > 92/92, suggesting that I'm also hitting up against a hard limit. The > WAP is connected to my main system via the motherboard ethernet, which > I know from other means (when I was transferring data from my previous > system to the current one) has no trouble maxing out at 10 MB/sec > (that ethernet does DMA). Not sure about that one; it may be that you're doing about as well as 802.11b and your access point will allow. The bit rate of 802.11b is 11Mbps, but there are other issues with wireless, such as latency (always higher than wired LANs) and delays for the collision avoidance.

April 2, 2002

Re: [Wlug] poor PCMCIA network performance

by Robert L Krawitz

Date: Mon, 1 Apr 2002 10:23:12 -0500 From: "Charles R. Anderson" <cra(a)wpi.edu> Using FTP to check the speed of the network is not a great idea, especially on slow laptop disks. You could be running into the limitations of your disk subsystem. There is a program called ttcp which will test TCP performance with nothing else involved. My laptop disk is fast enough (it's an IBM 32 GB 5400 RPM device); the disk streaming bandwidth (writing to a file) is in excess of 10 MB/sec. The disk on my main system is a SCSI disk that benchmarks at >30 MB/sec. Are those card 16-bit PCMCIA, or 32-bit Cardbus? If they are 16-bit, you are probably maxing out the bus. 16-bit PCMCIA is no better than the old ISA bus in terms of speed. 32-bit Cardbus is like PCI. The wired card is 16 bit. I don't know about the wireless Linksys; I suspect that it is also. For the 10/100 card, check the settings to be sure that both ends of the link are operating at the same duplex. If you are connecting to a repeater (hub, not a switch) you must be using half duplex. If you are connecting to a switch, then the card has to be configured for the same duplex as the switch or you will have horrible performance. Alternatively, the card and the switch can be configured for autonegotiation, so they will both agree on the "best" settings that they both support. Unfortunately, autonegotiation is sometimes buggy and sometimes doesn't work right between different vendor's devices. This is more of a problem with older network devices. I'll look into that. Thanks. -- Robert Krawitz <rlk(a)alum.mit.edu> http://www.tiac.net/users/rlk/ Tall Clubs International -- http://www.tall.org/ or 1-888-IM-TALL-2 Member of the League for Programming Freedom -- mail lpf(a)uunet.uu.net Project lead for Gimp Print/stp -- http://gimp-print.sourceforge.net "Linux doesn't dictate how I work, I dictate how Linux works." --Eric Crampton

April 1, 2002

T&G guotes

by doug waud

Hi I liked the following two quotes in today's (4/1/02) Telegram and Gazette (page A8) "Microsoft bids to acquire Catholic Church" and "Microsoft exploding an underground nuclear device" doug

April 1, 2002

Re: [Wlug] poor PCMCIA network performance

by Charles R. Anderson

Using FTP to check the speed of the network is not a great idea, especially on slow laptop disks. You could be running into the limitations of your disk subsystem. There is a program called ttcp which will test TCP performance with nothing else involved. Are those card 16-bit PCMCIA, or 32-bit Cardbus? If they are 16-bit, you are probably maxing out the bus. 16-bit PCMCIA is no better than the old ISA bus in terms of speed. 32-bit Cardbus is like PCI. For the 10/100 card, check the settings to be sure that both ends of the link are operating at the same duplex. If you are connecting to a repeater (hub, not a switch) you must be using half duplex. If you are connecting to a switch, then the card has to be configured for the same duplex as the switch or you will have horrible performance. Alternatively, the card and the switch can be configured for autonegotiation, so they will both agree on the "best" settings that they both support. Unfortunately, autonegotiation is sometimes buggy and sometimes doesn't work right between different vendor's devices. This is more of a problem with older network devices. On Mon, Apr 01, 2002 at 08:14:06AM -0500, Robert L Krawitz wrote: rlk> I use two network cards in my Dell Inspiron 8000 (PIII-700): a 3Com rlk> 3CCFE574BT and a Linksys wireless card. With the 3Com card (which is rlk> a 10/100) I never get better than 1.4 MB/sec (about 12 Mb/sec), even rlk> when the card is in 100 Mb mode (at some point in the past the limit rlk> was about 2.5 MB/sec). At that rate, top shows my CPU utilization rlk> pegged at about 100%, even doing a simple ftp. I know the card rlk> doesn't do DMA, but that's rather slow even for a PIO transfer. These rlk> cards have been around for ages; I don't think people would have been rlk> too satisfied with 400 KB/sec saturating a 200 MHz system. rlk> rlk> On the Linksys card, I likewise never get more than about 250 KB/sec, rlk> even if the computer is right next to the WAP and the link quality is rlk> 92/92, suggesting that I'm also hitting up against a hard limit. The rlk> WAP is connected to my main system via the motherboard ethernet, which rlk> I know from other means (when I was transferring data from my previous rlk> system to the current one) has no trouble maxing out at 10 MB/sec rlk> (that ethernet does DMA). -- Charles R. Anderson <cra(a)wpi.edu> / http://angus.ind.wpi.edu/~cra/ PGP Key ID: 49BB5886 Fingerprint: EBA3 A106 7C93 FA07 8E15 3AC2 C367 A0F9 49BB 5886

April 1, 2002

RE: [Wlug] T&G guotes

by Keller, Tim

Well, the first one I could actually conceive of... You figure, if they acquire the catholic church they'd just decree that as of baptism you signed the EULA (sorta like opening the package means you agree to the terms (that are on the slip of paper inside the package..)) since the EULA basically means microsoft could set off a nuclear device and get away with it, they'd have no problem proving in court that the church isn't responsible for faulty prietware. As for the nuclear device, If its a microsoft nuclear device I can see the following things happening... 1. Needs repeated reboots before exploding. 2. The trigger will need to have a service pack installed. 3. You'll need to buy MSND 1.5 to get the newest features (like exploding) 4. You'll need to have IE installed along with the latest direct X drivers. 5. before exploding you'll need to get a MSND registration number and enter it in binary using a telegraph pad. 6. "Explode" and "Disarm" will be the same button... Just some thoughts... Tim. -----Original Message----- From: doug waud [mailto:douglas.waud@umassmed.edu] Sent: Monday, April 01, 2002 10:36 AM To: wlug Subject: [Wlug] T&G guotes Hi I liked the following two quotes in today's (4/1/02) Telegram and Gazette (page A8) "Microsoft bids to acquire Catholic Church" and "Microsoft exploding an underground nuclear device" doug _______________________________________________ Wlug mailing list Wlug(a)mail.wlug.org http://mail.wlug.org/mailman/listinfo/wlug

April 1, 2002

poor PCMCIA network performance

by Robert L Krawitz

I use two network cards in my Dell Inspiron 8000 (PIII-700): a 3Com 3CCFE574BT and a Linksys wireless card. With the 3Com card (which is a 10/100) I never get better than 1.4 MB/sec (about 12 Mb/sec), even when the card is in 100 Mb mode (at some point in the past the limit was about 2.5 MB/sec). At that rate, top shows my CPU utilization pegged at about 100%, even doing a simple ftp. I know the card doesn't do DMA, but that's rather slow even for a PIO transfer. These cards have been around for ages; I don't think people would have been too satisfied with 400 KB/sec saturating a 200 MHz system. On the Linksys card, I likewise never get more than about 250 KB/sec, even if the computer is right next to the WAP and the link quality is 92/92, suggesting that I'm also hitting up against a hard limit. The WAP is connected to my main system via the motherboard ethernet, which I know from other means (when I was transferring data from my previous system to the current one) has no trouble maxing out at 10 MB/sec (that ethernet does DMA). eth1 IEEE 802.11-DS ESSID:"..." Nickname:"..." Mode:Managed Frequency:2.437GHz Access Point: 00:04:5A:CF:F3:B1 Bit Rate:11Mb/s Tx-Power=15 dBm Sensitivity:1/3 Retry min limit:8 RTS thr:off Fragment thr:off Power Management:off Link Quality:26/92 Signal level:-76 dBm Noise level:-149 dBm Rx invalid nwid:0 invalid crypt:0 invalid misc:0 I'm running SuSE 7.3 on both systems, with a recompiled kernel based on kernel-source-2.4.16.SuSE-24. I'm not (or don't think I am) running a firewall or NAT on the laptop, despite what's in the module list. I've had no luck finding anything on the net. Modules that look like they should be relevant are: orinoco_cs 4352 1 orinoco 27312 0 [orinoco_cs] hermes 3200 0 [orinoco_cs orinoco] af_packet 11616 0 (autoclean) appletalk 18832 0 (autoclean) ipx 15008 0 (autoclean) ds 6384 2 [orinoco_cs] yenta_socket 8368 2 pcmcia_core 37568 0 [orinoco_cs ds yenta_socket] ipv6 124480 -1 (autoclean) iptable_nat 12560 0 (autoclean) (unused) ip_conntrack 12656 1 (autoclean) [iptable_nat] iptable_filter 1728 0 (autoclean) (unused) ip_tables 10304 4 [iptable_nat iptable_filter] Suggestions? -- Robert Krawitz <rlk(a)alum.mit.edu> http://www.tiac.net/users/rlk/ Tall Clubs International -- http://www.tall.org/ or 1-888-IM-TALL-2 Member of the League for Programming Freedom -- mail lpf(a)uunet.uu.net Project lead for Gimp Print/stp -- http://gimp-print.sourceforge.net "Linux doesn't dictate how I work, I dictate how Linux works." --Eric Crampton

April 1, 2002

Re: [Wlug] Humorous signature file...

by Josh Huber

Chuck Homic <chuck(a)vvisions.com> writes: > Yeah, I think I know at least one person on the list with this > bumper sticker on his car... > > fsck! :) What a dork! ;) -- Josh Huber

March 30, 2002

RE: [Wlug] Request from academic researcher -- follow up

by Clint Moyer

Wow! I cannot even imagine a better example of misdirection... -----Original Message----- From: wlug-admin(a)mail.wlug.org [mailto:wlug-admin@mail.wlug.org]On Behalf Of Utpal M. Dholakia Sent: Friday, March 29, 2002 4:01 PM To: wlug(a)mail.wlug.org Subject: [Wlug] Request from academic researcher -- follow up Dear Linux User Group Moderator: I am a faculty member in the Jones school of management at Rice University in Houston, Texas (this may be verified using a Google search with "Paul Dholakia", or "Utpal Dholakia"). We are conducting an academic research project, studying social interactions online, and how communities organized around products or brands are formed, from a sociological perspective. In this regard, LINUX user groups and mailing lists provide a wonderful context to study these sociological issues. We are conducting a survey of members of LINUX user groups/ mailing lists seeking information on different elements of their thoughts, feelings, and motivations of participating in such groups, and the value of the group to them... The survey is available on the web at the following URL: http://www.surveypro.com/cgi-bin/surveypro/run_survey.cgi?id=2444 Of course, this research is strictly academic, and has no commercial links or interests whatsoever. We hope to publish this research in a sociology journal. As a small token of our appreciation for participants' time, and in the spirit of the philosophy underlying the open-source movement, we will donate $500 total to the five favorite organizations/ groups of our participants' choice (we ask the participant's to nominate their favorite groups in the survey). I am writing to seek your help in conducting this research, first through participating, and second, through encouraging your user group members to participate in the research.. This will not take more than ten minutes of your (and their time). Broad participation will increase the validity and generalizability of our findings, making the research more useful. We welcome all help, and look forward to hearing from you. Best wishes, Paul Dholakia Assistant Professor Rice University http://www.ruf.rice.edu/~dholakia _______________________________________________ Wlug mailing list Wlug(a)mail.wlug.org http://mail.wlug.org/mailman/listinfo/wlug

March 29, 2002

Request from academic researcher -- follow up

by Utpal M. Dholakia

Dear Linux User Group Moderator: I am a faculty member in the Jones school of management at Rice University in Houston, Texas (this may be verified using a Google search with "Paul Dholakia", or "Utpal Dholakia"). We are conducting an academic research project, studying social interactions online, and how communities organized around products or brands are formed, from a sociological perspective. In this regard, LINUX user groups and mailing lists provide a wonderful context to study these sociological issues. We are conducting a survey of members of LINUX user groups/ mailing lists seeking information on different elements of their thoughts, feelings, and motivations of participating in such groups, and the value of the group to them... The survey is available on the web at the following URL: http://www.surveypro.com/cgi-bin/surveypro/run_survey.cgi?id=2444 Of course, this research is strictly academic, and has no commercial links or interests whatsoever. We hope to publish this research in a sociology journal. As a small token of our appreciation for participants' time, and in the spirit of the philosophy underlying the open-source movement, we will donate $500 total to the five favorite organizations/ groups of our participants' choice (we ask the participant's to nominate their favorite groups in the survey). I am writing to seek your help in conducting this research, first through participating, and second, through encouraging your user group members to participate in the research.. This will not take more than ten minutes of your (and their time). Broad participation will increase the validity and generalizability of our findings, making the research more useful. We welcome all help, and look forward to hearing from you. Best wishes, Paul Dholakia Assistant Professor Rice University http://www.ruf.rice.edu/~dholakia

March 29, 2002

Re: [Wlug] Humorous signature file...

by Chuck Homic

On Fri, Mar 29, 2002 at 09:46:01AM -0500, Stephen C. Daukas wrote: > (The first day Microsoft produces something that doesn't suck will be the > day they start making vacuum-cleaners) Yeah, I think I know at least one person on the list with this bumper sticker on his car... fsck! :) -Chuck

March 29, 2002