(no subject)

From: Gerhard Schaden (ges@liscon.com)
Date: Tue Jul 01 2003 - 12:35:45 EST

Next message: Daniel Phillips: "Re: What to expect with the 2.6 VM"
Previous message: James Bottomley: "Re: [RFC] block layer support for DMA IOMMU bypass mode"
Next in thread: federicobriata@tiscali.it: "(no subject)"
Maybe reply: federicobriata@tiscali.it: "(no subject)"
Maybe reply: Zachary Pfeffer: "(no subject)"
Maybe reply: gigag@bezeqint.net: "(no subject)"
Maybe reply: Marcin Krasowski: "(no subject)"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Hi,

I've the following Setup

PC1
Linux PC kernel 2.4.21,
PCI Network card with realtek 8029

PC2
Windows 2k ( Terminal Server )
PCI Network card with realtek 8029
PCI Network card with realtek 8139

PC3
Windows 2k Terminal Server
PCI Network card with realtek 8139

Scenario 1:

PC1 and PC2 are connected via BNC and i use rdesktop to connect to the
terminal server, no problem
-------- --------
| PC 1 | | PC 2 |
|------| |------|
|-------------BNC--------------------|

Scenario 2:

PC2 gets a second Ethernet card rtl8139, routing is activated
PC3 is connected to PC2 via 100BaseT

-------- --------
| PC 1 | | PC 2 |-----100BaseT
|------| |------| |
    |-------------BNC--------------------| |
                                                      |
                                                  --------
                                                  | PC 3 |
                                                  |------|

i connect to PC3 with rdesktop no problem

Scenario 3:

The BNC Network gets a 3Com 8 Port Hub with one BNC

-------- --------
| PC 1 | | PC 2 |
|------| |------|
    |-------------BNC-----|--------------|
                          |
                        ----- --------
                        |HUB| -----10BaseT-----| PC 3 |
                        ----- |------|

PC1 connects to PC3 with rdesktop with collision rate around 30% with
network utilization 30KB/s (70pkts/s)

But if

PC2 connects to PC3 with mstsc I get the same network utilization but NO
collision

It looks like, that the Linux Networking stack is faster but produces
more collisions, I've looked at the throughput with iptraf and it was
perfect, even if i added more PCs making some traffic with tcpspray.
The problem is that customers don't accept the high collision rate.

I tried to change nearly all components:

Different Motherboards, different NICs (3Com), different HUB, different
Kernel (2.4.18, 2.2.20, 2.2.12), different cables :), different
T-Connectors, different Resistors

Tuning the software:
Limiting incoming traffic with iptables, limiting incoming traffic with
ingress scheduling,

Any Ideas ?

-GES

----------------------------------------------------------------
LISCON GmbH http://www.liscon.com/
Gerhard Schaden ges@liscon.com
Leitung Forschung u. Entwicklung

Europastrasse 8 Tel: +43 4242 214855
A-9500 Villach Fax: +43 4242 214855 99
AUSTRIA / EUROPE Mobil: +43 664 2333323

text/x-vcard attachment: Gerhard Schaden (E-Mail).vcf

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Next message: Daniel Phillips: "Re: What to expect with the 2.6 VM"
Previous message: James Bottomley: "Re: [RFC] block layer support for DMA IOMMU bypass mode"
Next in thread: federicobriata@tiscali.it: "(no subject)"
Maybe reply: federicobriata@tiscali.it: "(no subject)"
Maybe reply: Zachary Pfeffer: "(no subject)"
Maybe reply: gigag@bezeqint.net: "(no subject)"
Maybe reply: Marcin Krasowski: "(no subject)"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

This archive was generated by hypermail 2b29 : Mon Jul 07 2003 - 22:00:13 EST