Most of your basic switches are layer 2 only where as IP traffic from one device to the next on your network is layer 3. In this case, all traffic from one device to another on your network has to go out the sending device, through the switch, to the router to figure out where the receiving IP is, back through the switch and finally to the receiving device. You are limited by the throughput of your switch (keeping in mind the data goes through the switch twice) and the router (which is normally the bottleneck) as well as the read/write speeds of your hard drives (these usually run at about 150MB/s for a 7200RPM drive).
Hi I know I'm late to the party on this one but the above comment is only right if the two devices communicating aren't within the same subnet. The IP session is active between the NAS and the Computer, 'below that' there is an switched Ethernet session thru the switch in a pseudo-point-to-point fashion.
As for the speed issues you're seeing for the most part it is likely the speed of the hard disk, however (and to quote Ben Goldacre) "it's a little bit more complicated than that".... things to consider:
1 - Spinning hard disks are awful for writing data in an orderly fashion, they're spinning very quick and write data to the next available place. Imagine a teenager walking into their room and discarding their jacket in the first available floor space. This makes for sub-optimal reading and writing from day 1.
2 - The theoretical maximum data speeds of hard drive controllers are simply that... theoretical they aren't for the most part achievable in the real world. Hardware manufacturers place the HD controllers 'south' of mediation chips like southbridges which in turn don't have the facility to 'talk to the cpu' at even the speed the HDD controller abd HDD can sync.
3 - Network controllers tend also to 'live' off a southbridge (SB) or similar hence the data path from the disk to the network goes from the HDD controller, up thru the SB, gets processed, down thru the SB and on to the network. Classic data tromboning very inefficient and encountering multiple 'bottlenecks'.
4 - Ok finally out of the computer/NAS now, cheap/home network switches (whether they say Gigabit or not) are made cheaply, this often means that each Gigabit port on the switch shares the Gigabit speed to the switch 'backplane' with a cluster of other ports (for example in a Cisco Catalyst 3750v2 24 Gig port switch there is a 6-to-1 port-to-backplane ratio and they cost $6k upwards!). So more than likely the actual potential thru put of your switch is maybe a Gig divide by 4 or something similar, there's a saying "you get what you pay for" it's for the most part true with computer equipment too.
These are just things to consider, the jumbo framing idea should work as it increases the data payload to framing overhead ratio...
