The solution is simply a hardware upgrade for starters. A 32-bit/33 MHz bus
will top out around 200 Mbps.  If you look for a bus with higher speed
slots, you can triple your throughput without adjusting ANYTHING in your
Linux kernel.

Well thats where I disagree. And where I am looking for clarification.

We were NOT using systems with Buses limited to 32bit/33Mhz or Basic PCI.
The routers that we used for testing were using the Intel 7501 chipset, spec'd at allowing 3GB of throughput, and PCI-X Bus.

There are actually three potential limits...
1) Hardware
2) Software
3) Ethernet Protocol theoretical limit.

My understanding, although not confirmed, was that there was an Ethernet (cdma) timing Limitation, that prevented the Full 1GB from being reached using a < 1500MTU.

Myhardware can handle it. I'm investigating whether the limit is related to my Linux Software config versus Ethernet theoretical limits.

If you want wirespeed GigE performance with multiple packet streams at a
more reasonable packet size (remember that 1500 byte packets aren't
realistic), you'll need to make some adjustments to the kernel.

Well, thats my point... 1500MTU is a requirement that is usually beyond the control of the ISP. The ISP may may control the GB connection for Higher MTU, but NOT necessarily backend Transit or Front End last mile connections. The Internet is full of less than 384K average size packets, and not much an ISP can do about that.

Part of the question becomes, can near 1GB be acheived at 1500MTU?

With the 1500MTU frame acheiving only 200kbps, our routers CPU utilization
was less than 20%, so it was not a saturated router.

Actually, it is. Utilities top and vmstat don't necessarily reflect all of
the CPU utilization, and can't account for PCI bus contention/overhead.

Excellent Point. Any way to tell that (FULL CPU utilization)?

The second we changed MTU to 9600, we got over 800 mbps, and CPU
utilization was still very low, forget exact number but under 40%.

Well,'ve just taken your packet rate down significantly.

The reason we thought this was an Ethernet limitation and not a CPU/hardware limitation is that, we were able to pass a larger speed by combining multiple 100 mbps connections. For example, we were able to get 400 mbps with 4- 100mbps connections simultaneously.(didn't have more machines to test.)

still have bus overhead and the limitations of a 32/33 bus (1 Gbps burst

Again, we were using PCI-X not limited in that way, to my understanding.

Does Image Stream have any data, from live tests, proving the speed they can get across their GB routers at < 1500 MTU?

Maybe thats a solution for us?

Tom DeReggi



Tom DeReggi
RapidDSL & Wireless, Inc
IntAirNet- Fixed Wireless Broadband

WISPA Wireless List:



WISPA Wireless List:



WISPA Wireless List:



Reply via email to