Poor throughput from desktop to server

[Phantomaniac]

I have a FreeNAS server in my basement and my desktop (Win 7) upstairs. They are connected via Cat 6 cable and a TP-Link gigabit switch. I noticed my writes were somewhat slower than my reads (WD Black 1TB drives in RAID 1). Figured it was the drives but for shits and giggles I ran an internal test. Both my read and write are around 120MB/s, which should come close to maxing gigabit. So then I ran iperf and got some odd results:

From server to desktop: 900ish Mbps
From desktop to server: 600ish Mbps

The heck... I did some research, swapped cables on both ends, swapped ports in the switch, swapped NICs in my desktop (to Intel NIC from integrated Realtek), virtually nothing changed. Switching to Linux does give me about 50-100Mbps better throughput from desktop to server, but there is still at least a 200Mbps throughput discrepancy. The only thing I can think of now is to try a different switch. Any ideas? Does consumer grade network gear just suck?

 

[SecurityTheatre]

I have a FreeNAS server in my basement and my desktop (Win 7) upstairs. They are connected via Cat 6 cable and a TP-Link gigabit switch. I noticed my writes were somewhat slower than my reads (WD Black 1TB drives in RAID 1). Figured it was the drives but for shits and giggles I ran an internal test. Both my read and write are around 120MB/s, which should come close to maxing gigabit. So then I ran iperf and got some odd results:

From server to desktop: 900ish Mbps
From desktop to server: 600ish Mbps

The heck... I did some research, swapped cables on both ends, swapped ports in the switch, swapped NICs in my desktop (to Intel NIC from integrated Realtek), virtually nothing changed. Switching to Linux does give me about 50-100Mbps better throughput from desktop to server, but there is still at least a 200Mbps throughput discrepancy. The only thing I can think of now is to try a different switch. Any ideas? Does consumer grade network gear just suck?

Realistically, either it is due to the fact that your writes involve more random seek activity (try defrag), or because the network chipset in the NAS is slightly slower at writing, which is entirely plausible.

It's already way faster than most consumer-grade NAS solutions.

 

[Phantomaniac]

Realistically, either it is due to the fact that your writes involve more random seek activity (try defrag), or because the network chipset in the NAS is slightly slower at writing, which is entirely plausible.

It's already way faster than most consumer-grade NAS solutions.

I was using the iperf utility to test bandwidth, which should be entirely in memory and not utilize the disk. My server has a supermicro board with an Intel chipset, but you may be right. I'll try putting my external Intel NIC in the server and see if that flip flops my test results.

Most consumers aren't trying to boot from or host virtual machines via iSCSI. ;D