Storaxa kickstarter - an interesting small NAS

[PigLover]

The motherboard of the below kickstarter project looks like the CWWK board mentioned in this thread:

Fully Customizable Home Cloud Storage with Remote Access NAS

Linux OS | Max 142TB Storage | 5xHDD | 4xSSD | TrueNAS | WiFi 6 | 4x 2.5GbE | Intel CPU | SD+MicroSD Card Reader | 4x 10Gbps USB Ports

www.kickstarter.com

@newabc originally posted this in the long thread about the Topton fanless firewall boxes. But I think it is interesting enough to have its own thread (to make it more searchable).

Lots on compromises on this one but lots to like as well. The price for early-bird kickstarter supporters is really interesting for an N6005 based system. The 4x m.2 bay on top is very creative.

On the downside - those four m.2 drives are sitting behind an NVMe switch that is ultimately connected upstream with an x1 link (ugh!). Probably fast enough for NAS and certainly faster than SATA but not going to be amazing. And the extra m.2 on the motherboard is also x1 or x2 (unless they figured out some magic way to create extra PCIe lanes for the N6005).

Overall though its probably better than any other >$250 NAS chassis available today.

 

[newabc]

It is great with such design and low price.

I think my only concern is: if it uses the CWWK's design and only changed the 2 m2 nvme ports to a U.2 port, its 5 sata ports are from a JMB585 chip and I am not sure about its data transfer rate inside this JMB585.

I remembered someone bought the CWWK designed motherboard and complains its sata port's data speed in STH.

Update: The discussion thread link: https://forums.servethehome.com/index.php?threads/topton-nas-motherboard.37979/
Another N5095 board with 12 sata ports from Innovision was also mentioned on the 2nd page of this thread.

 

[newabc]

Another thing is that if it can only use non-ecc rams, there will be a concern with ZFS. Unbuffered ECC ddr4 ram is expensive too.

 

[Marsh]

My first backing of kickstarter project.
Like buying the board and get the case and power supply for free.

Overall though its probably better than any other >$250 NAS chassis available today.

 

[korikaze]

Other than form factor, it doesn't seem that great. They don't have the topology shown anywhere but mention each NIC gets 1 PCIe lane. The four NVMe slots share a single PCIe 3.0 x4 connection through U.2 which is pretty bad considering a single NVMe SSD can max that out. So that's the 8 CPU provided PCIe lanes maxed out already not even counting the 10Gbps USB ports. Is the chipset still getting 4 lanes on new Intel stuff? The six port SATA controller would be using 4 lanes for max throughput of each port if my head math is correct, though you wouldn't realistically get to even third of that with HDDs.

 

[PigLover]

Other than form factor, it doesn't seem that great. They don't have the topology shown anywhere but mention each NIC gets 1 PCIe lane. The four NVMe slots share a single PCIe 3.0 x4 connection through U.2 which is pretty bad considering a single NVMe SSD can max that out. So that's the 8 CPU provided PCIe lanes maxed out already not even counting the 10Gbps USB ports. Is the chipset still getting 4 lanes on new Intel stuff? The six port SATA controller would be using 4 lanes for max throughput of each port if my head math is correct, though you wouldn't realistically get to even third of that with HDDs.

its actually worse than you describe - but at the same time, reasonably fit for purpose as I'll note below.

• The four m.2 drives on top actually multiplex a single lane (yup - pcie x1 multiplexed across all four drives)
• The m.2 drive on the mb is also x1
• The SATA controller (Jmicron JMB585) is connected by a single x1 lane as well. It actually only takes 2 lanes to max out realistic 5x SATA3.
• The Sixth SATA port, the one they use for the boot drive, is the native SATA on the chip. But internally the N6005 SOC it is actually multiplexed with one of the PCIe lanes,
• Each NIC gets a single lane
• Above accounts for 7 of the 8 available PCIe lanes from the N6005. Fair to assume that the remaining x1 is used to provide the high speed USB

All that seems pretty bad. And if you were using the box for local processing and trying to maximize speeds from all the drives all at the same time it really would be pretty awful. Will it win any benchmarking competitions? Not even close.

But consider its purpose, it is built as a home/small office NAS with router options. No matter how fast your drives can go your ultimate limit for that application is how fast you move data on/off the box over the LAN. Can 5x SATA drives all keep up with and use most of a 2.5gbps LAN connection? Yup - most likely. Can an x1 connected m.2 drive saturate that same 2.5g link? Absolutely. Most home users will be linking over WiFi anyway, so cut those speed requirements way down again. And (not counting us wacko's on STH) the homes that are wired are almost universally 1g for the next few years at least. Can they saturate all four links at the same time? Probably not - but the "purpose" this was built for suggests those other links would be doing router functions anyway.

That is what I think makes this box interesting. As a small NAS it has really amazing capabilities. At $219 its competition are NAS boxes built using small ARM socs. Do the same analysis of speeds/feeds on those boxes and you'll see this will blow them out of the water. Most of those use USB multiplexed drive converter chips for their SATA ports and multiplex all of their SATA onto a single USB3.0 connection. A few of the better ones might have a single native SATA port and a port multiplier chip. This thing's x1 connected Jmicron controller is a speed demon by comparison. And I don't know of any in this price range offering any m.2 NVMe at all. So having 5 - even with crippled interconnect - is a slam dunk win.

Can a similar NAS be designed that is better, with x4 connectivity to all the m.2 and native (or at least full speed) SATA for all 5/6 drives? Yup. Absolutely. Can it be done for >$300? Not likely.

 

[PigLover]

One more note. Given the constraints of the N6005 PCIe configuration I think the only remaining design optimization for this would have been to put a mux chip behind the 2.5gbe connections and free up 2 lanes of PCIe, or even 3 lanes if you are willing to mildly oversubscribe the LAN. Statistically I think the chances of 4 links all maxing out together is pretty low so driving them all from 1 lane may be reasonable. Realistically, I'm not even sure why this box needs 4x 2.5gbe.

Then you could use the PCIe lanes to improve the interconnect to the m.2 array on top. Or you could properly connect the SATA controller chip at x2 and to give the m.2 array x3 (or the m.2 array x2 and the MB m.2 x2 as well).

This would require another PCIe mux chip so probably adds $10 to the end user cost.

But - as noted above - I actually think this box is already reasonably fit for purpose as a very low cost home NAS/Router.

 

[Haldi]

Hello,

Seeing as One PCI-E 3.0 Lane does 1Gybte/s you could saturate around 3 2.5GbE Ports with that.
As a NAS there are usecases where 4 people want to download Stuff from storage fast... if it's saved on NVME. via SATA HDD Raid you're not gonna get that 1.25Gbyte/s to max out 4x 2.5GbE ports.
But well... as all NVME Drives are stuck on 1 PCI-E Lane anyway i agree that multiplexing 1 PCI-E Lane to 4x LAN Ports should be enough.

But yeah. I like the idea of 4x 2.5GbE Ports... right now i don't have a 2.5Gbit switch here, and we do have 3 PC's right here that could all benefit from that speed. Mostly the 115-120MByte/s of the Connection are the issue while transfering files.

BTW they are talking about an Upgrade to Ryzen 7 5825U chip, which has 16 PCI-E Lanes (or are 4 Stuck on the PCH and you can only get 12?)

But i assume that would come in a 30-100$ price increase as seen in the Aliexpress Router HTPCs with 5825U vs N6005.



By how far they overshot their goal we might see a lof of interesting news in the next 30 days.



P.S
For the price of 219$+35$ shipping i also ordered one. Mainly to play around. Might even replace my HP Microserver Gen8 if it works well enough.

 

[mattventura]

This feels like a difficult niche in which to sell an enthusiast-grade product. In order to hit the price point, they have to load it up with consumer-grade hardware, chock-full of compromises as pointed out in this thread, yet you're trying to sell it to the crowd that's more than happy to grab some secondhand enterprise-grade gear and repurpose it into something that outclasses this both in terms of performance as well as features.

For example, 2.5GbE is certainly impressive to an average consumer who has been stuck on 1GbE for as long as they can remember, but it's a bit of a wet noodle when compared to cheap 10/40GbE stuff. Sure, it costs a little more upfront, but that will last for many generations to come.

 

[newabc]

It was designed as a home router + NAS that is why there are 4 2.5gbe ports on CWWK's original design. Actually if it is only used as NAS, the N5095 board with 12 sata ports is better for these designers to begin with.

 

[Meteor Man69]

I am not highly technical. More of a moderate geek personal user.

I am on the fence on supporting this Storaxa Cloud Storage device. I just purchased two OWC Thunderbay 8's and they work fine for local NAS. I just "upgraded" to the OWC's from DROBO 5D's because of the Mac OS compatibility concern going forward. (BTW my DROBO 5D's still work on my Mac using Ventura. I should have waited but did not want to take the chance of not being able to transfer my files.)

I like the idea of the Storaxa for backing up my iPhone and iPad and for using Time Machine for my Mac Studio and MacBook Pro. I am not thrilled with paying Apple for iCloud service ad infinitum.

Currently I only use the Thunderbay's in RAID 5 for files and data with one of them being a redundant clone back-up. The Storaxa would allow me to back-up my devices rather than just my data as I am currently set up.

The only problem I foresee is that all my drives are physically located in my home and if a disaster were to happen everything is lost anyhow.

Not sure I want to invest much more $$ at this point. In addition to the Storaxa pledge I would need to buy expensive HDD's (and maybe some SDD's) to populate the Storaxa. Luckily for my OWC Thunderbay's I had a bunch of HDD's to fill them with.

What are others doing? Data? Device Back-ups? Redundancy and Emergency plans?

 

[BoredSysadmin]

Storaxa would be running Truenas scale. It's not the most complicated storage OS, but not easiest one either.
Some users were successful using TS with the TimeMachine backups:
https://www.reddit.com/r/truenas/comments/usmuyk

 

[Haldi]

What are others doing? Data? Device Back-ups? Redundancy and Emergency plans?

Put my old NAS at my Parents place....
offshore backup


P.S
they have a specially modified Mainboard made for them.



And MoQ for the AMD Ryzen 5825U with ECC Support would be 500

 

[ano]

really needs sfp+ or 10G-base-t to be usefull.

 

[BoredSysadmin]

really needs sfp+ or 10G-base-t to be usefull.

not in this design. It's severely limited by a lack of PCIe lanes.

 

[Unrest5078]

Had a great day today.
I searched the net about n6005, landed on your awesomely(is this even a word) based forum, found this catchy project, made my first ever pledge on KS and my 1st post here. I, hereby thank you! I'm with the big guys' league
(coming from a fractal 304+ asrock j5005 which I needed to upgrade due to lack of sata).

 

[Haldi]

 

[Unrest5078]

As excited as I've been these days, slowly it becoming more and more clear this is an elaborate scam. As some other pointed out, it's close to impossible to be profitable with such a build, given the parts are so scarce at the moment. Not to mention the recent cooking "upgrades" (ryzen 7).
I withdrew my backing, and advise others to give it more thought, if they already backed it.

What are your thoughts?

Some comments from KS

 

[Haldi]

As excited as I've been these days, slowly it becoming more and more clear this is an elaborate scam. As some other pointed out, it's close to impossible to be profitable with such a build, given the parts are so scarce at the moment. Not to mention the recent cooking "upgrades" (ryzen 7).


What are your thoughts?

I wouldn't call it a scam outright.
Rather fishy.
But I do think it's technically possible to make it feasible in that price!
First off we're not buying 5 or 10 devices but over 500. That will greatly reduce costs.
Second. Prices are not going up but down. Only GPU prices are increasing all other electronics are slowly going downwards and are supposed to go even further down in the Future.
What if some supplier is sitting on a huge stock of N6005 CPUs he can't get rid of? Next Generation has already been released so it's old stuff.
Also maybe they have a special deal. Get a fixprice now for hardware they will get in 3 months. And the supplier is gambling on the falling hardware prices.
RAM isn't really an issue. Refurbished RAM from Server farms or Mining Rigs still works pretty well and doesn't cost much.
Also the 128gb SSD is probably around 5$ to 15$ at most.

Now the real question is how do they pay the work.
Assuming one assembly and packaging takes about 30 minutes. Installing and testing software another 15 minutes that would be 45 Minutes per Unit.
Okay maybe they're WAY more efficient and can do all together in 20 minutes. That's still a LOT of time for the planned 2'000 Units.
I have no clue how much you'd have to pay for a "decently" qualified worker in China per Hour but I assume it's not too much. Minimum wage is 1.52$ per hour.



The real Question should be: are those people Trustworthy?
And we can't answer that.
Do you know of anyone who can read Chinese?
Maybe a quick Background check of these guys will shed some new light onto this Situation.

Or anyone knows of a private investigator in China that we could hire? I'm certain there are more people interested in that so via Crowdfunding we could easily get enought for an in dept investigation.

 

[Unrest5078]

What if some supplier is sitting on a huge stock of N6005 CPUs he can't get rid of?

That's pretty impossible (the getting rid of part), since already prices on 5105 have been going up by the week and 6005 are already becoming less and less available (sellers with both cpu options)

Unfortunately, after digging some more about prices and availability, for me, the answer is clear.
You cannot (at least profitably) sell such a product (not to mention the addons, shipping perks etc) and meet your deadline/goal.
I was in "X files" myself for a bit. I want to believe.
Not anymore. I really want this project to succeed and prove me wrong though, but I'm short on cash and the risk to reward ratio is too high for my taste.