 You don't need to be an 'investor' to invest in Singletrack: 6 days left: 95% of target - Find out more
Can any networking gurus shed any light? I downloaded and ran some speed test software to see what the difference in performance was between devices plugged into our switch which has two gigabit ports (the rest being 100Mbps). The software basically just generates a file and times how long it takes to write it to another computer on the LAN and then reads it back.
On the two computers on the Gigabit connection it's reading at 700-800 Mbps (as expected) but writing at only about 60 Mbps. Is this normal or do things need tweaking? The result is the same running the software on the other Gigabit machine (slow upload, fast download).
On other machines without the Gigabit connection there is a difference but it's less dramatic (e.g. 90 Mbps read/35 Mbps write). Should the write speed ideally be closer to the read speed or is writing always going to be slower?
All these machines are Macs with SSDs.
What software are you using? You've got the throughput of the disk subsystems as well as the network if you are reading and writing a file.
Lperf will give you memory to memory writes if you want to cut out the disk.
Edit: Sorry. Typo. As pointed out, iperf.
Try iperf.
Considering the following article, and as you're movinng things between SSD's, that maybe seems a mite slow:
http://www.tomshardware.com/reviews/gigabit-ethernet-bandwidth,2321-7.html
Can't help with the why though unfortunately.
The free version of LAN Speed Test from totusoft.comWhat software are you using?
Just the first thing that popped up when I googled!
Installed and ran iperf on the two gigabit machines, first using one as the server and one the client then swapping round and running test again. Both times averaging about 940 Mbp/s i.e. basically full speed.
So I guess that means it's the disk write operation that's slowing things down. Just got to find out why now!
Are you interpreting Mb and MB correctly?
I notice you mention Macs, there's your problem. You need to upgrade immediately to the latest.
I understand there is now a Gnarmac..... I'll get my coat.
The software could be bollocks to be fair.
Simplest test is just to copy a big file and see how long it takes. mkfile will create you a file, and I think rsync can give you stats (don't have a Mac near me to test it out).
Do you just want to benchmark, or is there a particular performance criteria you need to meet for an application?
The software could be bollocks to be fair.
That was my first thought. Or rather, "do we trust the software?" I'd be getting a second opinion from something else before entertaining the idea of frobbing the network settings.
I always used ttcp, no file writing going on...
You need to break the test down and test:
1. LAN speed on it's own without any files eg iPerf, NetPerf etc
2. Separately test disc/SSD read / write speeds on the local machine
Have a gold star! You too Cougar. Yes, if I just copy a big file over it performs as expected i.e. approx 10s for 1GB on the gigabit connections and 10x that long for the others.The software could be bollocks to be fair.
Just want to make things as fast as possible really as we routinely deal with fairly large files. I thought the speed-test software would help to see what the difference was between machines connected with gigabit ethernet and those not. Looks like just copying some files over and timing them on my phone was a better idea though!Do you just want to benchmark, or is there a particular performance criteria you need to meet for an application?