What would be an expected transfer through a Gig-E switch?

[Texun]

I recently purchased a Netgear 5 port gig-e switch and have it connected to two of the four PC's in my apartment (silly, I know). The speeds are WAY better than with the 10\100 router but I'm not sure exactly what to expect.

Using this conversion page

my speeds were around 18MBytes (144mb) max between PCs when I first fired it up. I enabled flow control in my adapters and now peak at about 30MB (240mb). Average is 20-25. My MTU is still at 1500.

I know there's overhead, but is this a normal speed?

One PC has an onboard nVidia gig-e and the other has a USR gig-e NIC. I'm using CAT5E.

Thanks

EDIT********

More info: I am transferring media files; images, mp3 and some DVD backups, usually 1 or more gigs at a time - fairly well compressed files. CPU utilization shows 40% average during the transfer.

Thanks

 

[phatrabt]

My understanding is that with GigE you're usually going to be limited by your drives. I could be wrong though.

 

[Texun]

Of course..... now that makes sense. Thanks.

 

[Genx87]

Sounds about right, on my GigE network at home, I usually avg about 20-25Mbps when writing or reading from my discs.

 

[JackMDS]

Home Giga Network: http://www.ezlan.net/giga.html

Giga networking - http://www.ezlan.net/giga_net.html

 

[robmurphy]

If the NICs on the PCs support it then enable the jumbo frames. You will need to use the lowest, so if one PC supports 7K and another supports 9k you will need to use 7k. This may cause problems if the gigabit network is used to connect to the internet, so be you may have to go back to the 1500 MTU.

Rob Murphy.

 

[Madwand1]

Originally posted by: Texun
Using this conversion page

If you don't use legacy notation (i.e. do use M=10^6), you get a brownie point and a ~5% numeric boost.

http://en.wikipedia.org/wiki/Megabyte
http://en.wikipedia.org/wiki/Mebibyte

~30 MB/s is typical for single drive transfers among random machines. You can do a lot better in some cases, but these are valid typical numbers.

The switch has little to do with this performance -- many budget switches can hit "wire speed". There are many other bottlenecks, including the NIC settings (as you've seen with flow control options), OS, OS tuning, hard drives, and file transfer protocol. NICs themselves can be bottlenecks, but IMO these are not frequently seen at the single drive level.

Direction of transfer also makes a difference -- push is generally faster than pulls.

With the right setup, you could see close to drive speed for transfers -- this varies according to the type and condition of both drives involved. Empty drives perform the best, near-full drives the worst. Newer drives tend to performance better than older drivers, etc..

 

[Texun]

Originally posted by: Madwand1

Originally posted by: Texun
Using this conversion page

If you don't use legacy notation (i.e. do use M=10^6), you get a brownie point and a ~5% numeric boost.

http://en.wikipedia.org/wiki/Megabyte
http://en.wikipedia.org/wiki/Mebibyte

~30 MB/s is typical for single drive transfers among random machines. You can do a lot better in some cases, but these are valid typical numbers.

The switch has little to do with this performance -- many budget switches can hit "wire speed". There are many other bottlenecks, including the NIC settings (as you've seen with flow control options), OS, OS tuning, hard drives, and file transfer protocol. NICs themselves can be bottlenecks, but IMO these are not frequently seen at the single drive level.

Direction of transfer also makes a difference -- push is generally faster than pulls.

With the right setup, you could see close to drive speed for transfers -- this varies according to the type and condition of both drives involved. Empty drives perform the best, near-full drives the worst. Newer drives tend to performance better than older drivers, etc..

This is a detailed and very informative response! Many thanks to all. :beer: :beer: :beer: :beer: :beer: