efficiency of Ethernet?

[YKhan]

Well, I got a laptop with both a Wi-Fi and an Ethernet port on it.
Usually I'm using Wi-Fi unless I need to do something really intensive
between two computers within my home network, such as printing or file
transfer when I switchover to Ethernet. Now although 100BT Ethernet is
loads faster than WiFi, it just doesn't seem fast enough. It takes me
almost exactly 10 minutes to transfer a CD's worth of data (700MB)
between two nodes in the network, I should expect that to be done
within 1 minute at 100% efficiency. But I'm seeing only about 12.5%
efficiency out of the Ethernet. I had previously assumed that
efficiency might be around 60-70%, but I was surprised to only see this
level of efficiency.

I've tried this experiment both through the switch of a broadband
router, as well as a direct computer-to-computer X-over Ethernet cable.
It was the same in both cases.

Yousuf Khan

[daytripper]

On 20 Jan 2005 11:54:53 -0800, "YKhan" <yjkhan@gmail.com> wrote:
Well, I got a laptop with both a Wi-Fi and an Ethernet port on it.Usually I'm using Wi-Fi unless I need to do something really intensivebetween two computers within my home network, such as printing or filetransfer when I switchover to Ethernet. Now although 100BT Ethernet isloads faster than WiFi, it just doesn't seem fast enough. It takes mealmost exactly 10 minutes to transfer a CD's worth of data (700MB)between two nodes in the network, I should expect that to be donewithin 1 minute at 100% efficiency. But I'm seeing only about 12.5%efficiency out of the Ethernet. I had previously assumed thatefficiency might be around 60-70%, but I was surprised to only see thislevel of efficiency.I've tried this experiment both through the switch of a broadbandrouter, as well as a direct computer-to-computer X-over Ethernet cable.It was the same in both cases. Yousuf Khan

1. All network host adapters are not created equal.
2. Same thing goes for drivers.
3. And you still need enough horsepower to drive 'em fast...

/daytripper

[Bob Willard]

YKhan wrote: Well, I got a laptop with both a Wi-Fi and an Ethernet port on it. Usually I'm using Wi-Fi unless I need to do something really intensive between two computers within my home network, such as printing or file transfer when I switchover to Ethernet. Now although 100BT Ethernet is loads faster than WiFi, it just doesn't seem fast enough. It takes me almost exactly 10 minutes to transfer a CD's worth of data (700MB) between two nodes in the network, I should expect that to be done within 1 minute at 100% efficiency. But I'm seeing only about 12.5% efficiency out of the Ethernet. I had previously assumed that efficiency might be around 60-70%, but I was surprised to only see this level of efficiency. I've tried this experiment both through the switch of a broadband router, as well as a direct computer-to-computer X-over Ethernet cable. It was the same in both cases. Yousuf Khan

The efficiency of Ethernet is great as long as collisions are rare (as they
will be on a Xover cable with both NICs in FDX mode). But you are likely
not measuring the performance of Ethernet, but rather of a long HD-HD chain
involving great steaming wads of M$ (I assume) code; and, the Ethernet and
its NICs are probably not the bottlenecks.

For starters, check that your NICs are in FDX 100 Mb/s mode, with auto-
negotiation turned off.
--
Cheers, Bob

[Yousuf Khan]

daytripper wrote: 1. All network host adapters are not created equal. 2. Same thing goes for drivers. 3. And you still need enough horsepower to drive 'em fast...

Well, how fast have you seen it go?

Yousuf Khan

[Yousuf Khan]

Bob Willard wrote: For starters, check that your NICs are in FDX 100 Mb/s mode, with auto- negotiation turned off.

We've tried that too, i.e. turning auto-neg off and hard-coding the
connection scheme. It's still the same.

Yousuf Khan

[CJT]

Yousuf Khan wrote: daytripper wrote: 1. All network host adapters are not created equal. 2. Same thing goes for drivers. 3. And you still need enough horsepower to drive 'em fast... Well, how fast have you seen it go? Yousuf Khan

Moments ago I moved 1073739776 bytes via FTP in 140 seconds, for
a rate of 7.7 MB/sec (as reported by FTP). That was nothing special --
and it's probably more an indication of the speed of the disks and CPUs
involved than it is of any shortcoming of Ethernet. (FWIW, the two
machines involved have 333 MHz clocks.)

Still, at 8 bits/byte (ignoring overhead), that works out to an
efficiency of about 61.6%.

--
The e-mail address in our reply-to line is reversed in an attempt to
minimize spam. Our true address is of the form che...@prodigy.net.

[CJT]

CJT wrote: Yousuf Khan wrote: daytripper wrote: 1. All network host adapters are not created equal. 2. Same thing goes for drivers. 3. And you still need enough horsepower to drive 'em fast... Well, how fast have you seen it go? Yousuf Khan Moments ago I moved 1073739776 bytes via FTP in 140 seconds, for a rate of 7.7 MB/sec (as reported by FTP). That was nothing special -- and it's probably more an indication of the speed of the disks and CPUs involved than it is of any shortcoming of Ethernet. (FWIW, the two machines involved have 333 MHz clocks.) Still, at 8 bits/byte (ignoring overhead), that works out to an efficiency of about 61.6%.

BTW, I should probably also note that the receiving machine was fairly
heavily loaded doing another task at the time.

--
The e-mail address in our reply-to line is reversed in an attempt to
minimize spam. Our true address is of the form che...@prodigy.net.

[Yousuf Khan]

CJT wrote: Moments ago I moved 1073739776 bytes via FTP in 140 seconds, for a rate of 7.7 MB/sec (as reported by FTP). That was nothing special -- and it's probably more an indication of the speed of the disks and CPUs involved than it is of any shortcoming of Ethernet. (FWIW, the two machines involved have 333 MHz clocks.) Still, at 8 bits/byte (ignoring overhead), that works out to an efficiency of about 61.6%.

60% is what I used to see on my Unix servers at work. That's why I came
to expect and assume an efficiency of 60-70% as standard on 100BT. Come
to think of it, I used to measure their efficiency with FTP transfers
too, but I'm measuring these machines using Windows' file and printer
sharing. Let me try a couple of tests with FTP and see what happens
between these machines.

Yousuf Khan