Are you sure you got the facts right ? Don't seems to be consistent with my knowledge
A normal Cat 5E operating at at least 250MHz would be good enough for gigabit purpose. Cat 6 is not necessary unless you have a large distance to cover or your area is filled with tons of radiation interferences. Normal consumer home shouldn't be of such issue unless you place your network cable near to power sources or power lines. CAT 6 are especially good in data centre due to the large number of electronics in the premises.
First speed is subjective to purpose. Obviously it is recognized that consumer grade NAS are not going to match up with dedicated Enterprising solution, but I don't think you need a 3K grade system to get good performance that surpass most consumer.
I got my Synology DS1511+, which is a SOHO class system with 5 bays only cost me 1.5K. With 3x 2T Green WD drives, each only $130 in today's price, means a total of $390, I can already exceed 100MB/s for write and read using consumer level L2 switches sequentially consistency at above 80MB/s. As for Random I/O access, I am not going to say it's fantastic.
If I will to spend S$200 for each 300GB WD Velociraptor, I can get 5 of them for 1K. Using RAID 5, I can get 1200 Million Bytes worth of disk space, roughly slightly less than 1.1TB worth of actual diskspace. I can tell you 5x WD is going to pack quite sufficient Random I/O, though not as good as SSD for sure.
However, it's still not even 3K !!!! How did you come up with such imaginary number ?
It's true, moving 1TB at sustain 80MB/s is going to take 3.6hours. Can I ask if you transfer 1TB internally using SATA, is it going to be significantly faster ? The answer is NO.
Reason very simple, Not to mention SATA 6G, just use SATA 3G as the example. Having a faster interface is not going to make your actual throughput of your harddisk faster it is still going to sustain at roughly between 80-100MB/s. As such the argument that because it takes a long time to transfer 1TB from the NAS to the COMPUTER and hence it makes a bad choice to use NAS is totally nonsense. Even internally, 1TB of data to be transfer from 1 harddisk to another is going to take almost the same time, probably just marginal better due some factors here and there. But using this example to call NAS unsuitable, I don't know where you learn, but definitely not from what I experienced.
There are a few important points required to iron out. NAS doesn't require high processing power, but rather good implementation of the actual network filesystem used, and good operating system, good network interfaces with high processing throughput. Memory modestly provisioned will do. The only time where you need more is when you are also running something else on those NAS with bells and whistles. Data Deduplication, if supported at all, also require substantially more memory.
Next is if performance is paramount, then iSCSI should be used instead of NFS, or CIFS, or AFP. Not all NAS have good iSCSI target implementation. Format the iSCSI block device to the required filesystem used by the host. Ensure the network connectivity is good between the NAS and the host. That should give rather good results.
Conclusion, I find your explanation about NAS rather ill informed and not facts.
If really wanted a good NAS, one can DIY one. It will not be as feature rich as those like QNAP, Synology, Thecus and so forth, but it will serve the purpose of a dedicate NAS and still don't require 3K. I suspect you can even add it a 64GB SSD for read and write-back caching. All these using OpenIndiana(ZFS) on RAID-Z with L2ARC and ZIL.
