[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: OT: Maximum number of fast ethernet cards

Hash: SHA1

Pete Clarke wrote:
> Hi there,
> This is slightly offtopic, but does anyone know the maximum 
> number of fast ethernet cards a typical PC can handle..? I want 
> to use a cheap (ish) Debian box as a firewall/router to 
> suppliment my Netgear, and provide more services to the internal 
> network than currently available.

Like what?

> With this in mind, I was intending on slapping a load of dual 
> portal network cards into a redundant PC. I remember reading a 
> while back that the ultimate limit will be the PCI bus bandwidth 
> - with this in mind, what would be the maximum number of NIC's I 
> could realistically install into a PC?

That's right.  A bus can't carry more than the bus can carry...

For bog-standard 33MHz/32bit PCI, that's 133MBps theoretical which
is shared with the IDE controller.

Fast Ethernet is 100Mbps == 12.5MBps, which means you could
supposedly fit in 10 FastE *ports* before flooding the bus.  But
this will only give you enough extra bandwidth to write log files
out to disk.

Since no one ever gets the theoretical maximum, lets drop that down
20% to 8 ports.  Which means 4 dual-port cards.  Doesn't look like a
problem for a full-sized ATX mobo.  But then, when have you ever
seen Ethernet running a full capacity?  So, 10 ports *may* work.
I'd stick with 8 ports, though.

PCI-x cards (64 bit and/or 66MHz) will arithmetically increase the
bandwidth.  Don't know the bandwidth of PCI-e, but you can find it
in Wikipedia and do the math (sums? reckoning?) yourself.

Regarding CPU, modern CPUs are *fast* in text-only mode.  No,
*stupendously* fast.  A Pentium 90 *should* be able to handle that
much data.

The *definite* problem I foresee is interrupts.  Every 1500 byte
packet generates a hard interrupt, so that's a theoretical max of
8738 hard interrupts per port.  With 8 ports, that's 70,000 hi's per
second.  Can the Linux kernel handle that on a given bit of HW?
You've just flown beyond my area of expertise...

- --
Ron Johnson, Jr.
Jefferson LA  USA

Is "common sense" really valid?
For example, it is "common sense" to white-power racists that
whites are superior to blacks, and that those with brown skins
are mud people.
However, that "common sense" is obviously wrong.
Version: GnuPG v1.4.3 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org


Reply to: