Whilst your values are obviously lower than theoretical maximum I stated (and lower than I consider average performance), they also obviously back up my claim
The numbers you just stated for YOUR system doesn't get remotely close to the numbers you originally claimed and even giving you doubling your speed by adding two drives, that's still putting you WAY under the speed you originally claimed of 800-900MB/s for a 6 drive RaidZ2 (which would be 200-225MB/s per drive). In fact, that puts you closer to my numbers than your original claims. Ignoring my experience, I can't find anyone getting close to the numbers you're claiming when running valid benchmarks on a RaidZ2 setup, even using the newer high capacity WD Red's which are among the fastest of the consumer spindles. Fastest benchmark's I've seen of the newer large WD Red's puts them at 180MB/s best case scenario. I see multiple benchmarks on the older Red's, including here on AT that put them at 120-140MB/s. The large bulk of consumer drives are in that same range. Those are all single drive numbers and large file transfers, meaning best case scenario. You're not going to get peak per-drive performance in a RaidZ2 array. Therefore the AVERAGE RaidZ2 setup isn't coming close to your claimed numbers without other assistance (IE caching) so I'm not sure what your "average" number are based on.
A quick Google search of people running 6 drive RaidZ2 arrays, 430-480MB/s was the pretty consistent result, even with WD Red's. Keep in mind, RaidZ2 is hardly your best option (even within the ZFS world) if you're that concerned about performance and it's REALLY easy to get artificially inflated numbers on ZFS rigs. For example:
Code:
write 20.48 GB via dd, please wait...
time dd if=/dev/zero of=/tank/dd.tst bs=2048000 count=10000
10000+0 records in
10000+0 records out
real 6.5
user 0.0
sys 6.4
20.48 GB in 6.5s = 3150.77 MB/s Write
wait 40 s
read 20.48 GB via dd, please wait...
time dd if=/tank/dd.tst of=/dev/null bs=2048000
10000+0 records in
10000+0 records out
real 3.8
user 0.0
sys 3.8
20.48 GB in 3.8s = 5389.47 MB/s Read
That's off one of the two ZFS boxes I'm running right now at my house. If you've got some actual data showing a 6 drive RaidZ2 benching at 800-900MB/s without "cheating", by all means feel free to post it. So far your numbers look and sound like pure speculation from somebody who just jumped on the ZFS bandwagon. Not to mention these are direct to drive benchmarks and do not reflect the performance you'll be getting on network transfers.
5Gb/s theoretical max being almost 600MB/s, but reality will be closer to 500. So why are you pretending to offer a counter-point?
Because you completely ignored the rest of my post. The number of people with network storage systems in their home that can saturate GbE is
extremely small and most of them (like myself) have enough money into it that spending $300-$500 for high speed connectivity using widely available surplus equipment is a non-issue. Regardless if you're getting your numbers or my numbers for a RaidZ2 array, that does not reflect an average consumer setup. If you're running a "normal" 1 PC to 1 NAS setup, you can do 40Gb/s Infiniband for like $150. Have more systems than that? A 3 PC 40Gb/s Infiniband setup will run you $500-ish. Need longer distances but don't want to break the bank? 4Gb/s FC setups can be done for under $150. I wouldn't hold my breath on being able to build a 802.3bz for less than that in the next 5 years because there simply isn't enough demand for it.
I'm all for faster speeds but the number of people who can make use of faster than gigabit network speeds in the home is extremely small. You used the term "NAS Geeks". Most "NAS geeks" already have faster than 1GbE network speeds and aren't running RaidZ2 any more.