>On Mon, 28 Feb 2000, jamal wrote:
>
>> I _have_ measured the number of interupts per second that a PC could
>> withstand running Linux; This is very recent.
>> I was able to get something around 90K/sec when forwarding without utterly
>> slowing down the system; At 64byte packets thats close to 100Mbps; at
>> 640Byte packets thats close to 1Gigabit. This is of course without turning
>> on Alexeys fast forwarding core. I think at some point the PCI bus becomes
>> your bottleneck.
>
> Raw PCI 2.1 bus in a PC is 1056Mbps (32bit, 33Mhz) minus
>overhead. IIRC the Cisco 7206VXR is rated for 600Mbit *max* sustained
>traffic over its main bus (33Mhz 32bit PCI) given the 600 - 800 Mbps a
>high end linux system can do. (Remember though, shared HD bus, this
>is *total* transfer, not network transfers each way)
Exactly. The 7206VXR will still outperform a high end linux system,
because there are other demands on the bandwidth of the PCI bus in your
linux system. Also, it's worth noting that while Cisco classifies the 7200
series in their 'High-End Routers', it's a cheap box intended for
enterprise work. The newer 7576 can push 4Gbps sustained across the
chassis, and even the older 7507/7513 can do 2Gbps. I'm not suggesting
that we'll ever get linux to touch the capabilities of a 12000 (or a
Juniper M20/M40), but PC hardware issues aside, the kernel ought to be able
to push packets as well as any Cisco 7xxx series router.
As you've shown, obviously some PC architectures (AXP and PPC, for example)
are already to the point of pushing 2-4Gbps across the bus. In these
cases, bus bandwidth shouldn't be holding anyone back in the enterprise
routing arena (as long as those architectures can actually push 2-4Gbps
across the bus, as opposed to just being 'capable' of it).
--
Nick Bastin
Software Developer
OPNET Technologies
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
