Do not try this at home: Man spends $5,000 on a 48TB Raspberry Pi storage server

YouTuber tech whizz Jeff Geerling has found it is possible to spend $5,000 on a Raspberry Pi build. He did this silliness because he's a self-described "creative person who builds great software" with a sponsorship from Lambda that keeps him from redirecting his own paycheck to quirky projects. He's also undoubtedly …

[Anonymous Coward]

Coincidence

I found his blog last night(*) - it became apparent quite quickly he wasn't so much interested in solutions, more "how far can I push this before it gets silly". A true engineer.

* while hunting for a Raspberry Pi on a PCIe card, not with a PCIe card - to use as a lights-out controller. No such thing apparently.

[Mike 137]

'"how far can I push this before it gets silly". A true engineer'

With respect from a chartered systems engineer, that's not a typical (or even desirable) attribute of engineers, whose duty is to provide solutions to real problems that are effective, reliable, convenient and efficient.

It approximates closer to the approach of wizards like Wozniak, who are innovators rather than engineers. And we need innovators, but their creations seldom stand any long term test as engineering. The engineers must take over once the innovators have delivered, to refine the concept into a product.

[Totally not a Cylon]

Re: '"how far can I push this before it gets silly". A true engineer'

A good Engineer should be both.

Able to build something exactly to spec....

And to look at something like a Raspberry Pi and go "I wonder if....."

Stevenson, Brunel, Watt and the rest all pushed the boundaries of what was possible.

An Engineer is someone who works with engines/structures.

Prototypes are blue-sky thinking and should make people go 'Wow, I did not think that was possible/sane'

Whereas Production should be accurate to the design.

Part of Today's problems is manglement don't know the difference......

[Mike 137]

Re: '"how far can I push this before it gets silly". A true engineer'

"An Engineer is someone who works with engines/structures"

This is a common misapprehension. "Engineer" derives from the same root semantically as "ingenious", not from "engine". Indeed "engine" drives independently from the same root.

This misapprehension has caused every technician and fitter to be miscalled "engineer", and has thus caused great confusion.

In reality, an engineer typically converts ideas and concepts into specifications of realisable deliverables via processes of planning and design, then passes those specifications to technicians to be implemented.

On the other hand, innovators are obliged to undertake the entire end to end process, specifically because everything from the idea to the implementation is likely to be novel. Consequently, the results will commonly be prototypes, rather than finished products.

[Anonymous Coward]

Re: '"how far can I push this before it gets silly". A true engineer'

@Mike 137

Have upvoted you so I can call you a swot without feeling guilty.

SWOT

[teebie]

Re: '"how far can I push this before it gets silly". A true engineer'

'This misapprehension has caused every technician and fitter to be miscalled "engineer"'

...and a helpdesk full of "customer solution engineers".

[cb7]

Re: '"how far can I push this before it gets silly". A true engineer'

"Part of Today's problems is manglement don't know the difference......"

That's because manglement also have to listen to the bean counters or they'll be out of a job.

They find it's easier to say no than have the hard conversation that says we need more money so we can invest in R&D.

But that involves working out a ROI (Return on Investment) at which point most give up and carry on with business "as normal".

[Anonymous Coward]

Re: '"how far can I push this before it gets silly". A true engineer'

And part of the engineer's role is to know the limits of the thing they're working with, so they can make sure that the design doesn't reach them.

For some things those limits are well known (eg tensile strength of a specific wire), but in other cases you may have to test to destruction to find those limits. And then design to a maximum of half that limit.

But in this instance, anyone with any Pi experience would be able to tell you that the throughput is far too small for that type of use. You don't have to build it to know that.

[K-Tat]

Re: '"how far can I push this before it gets silly". A true engineer'

"But in this instance, anyone with any Pi experience would be able to tell you that the throughput is far too small for that type of use. You don't have to build it to know that."

Similarly, anyone with any internet experience knows "read the post"/"watch the video" before commenting.

He literally spends several minutes talking about "while you can do this, there's really no reason other than curiosity to do so" adding "the processor is a huge bottleneck and is unable to deliver at full speeds. He goes on to test it in several configurations that highlight this fact.

Good for you, proudly discovering something that was right in the video.

[Alan Brown]

Re: '"how far can I push this before it gets silly". A true engineer'

Hypothesising that something is incapable of performing a task well is not the same as actually demonstrating it in practice

I've been involved in a number of VERY expensive projects over the years where I could see issues building and was pooh-poohed upon raising doubts that the system would cope under full load, only for the wheels to fall off very spectacularly

I've seldom (if ever) been patted on the back and congratulated for catching something before it happened (more often vilified, even when I've started already making alternative plans to get the organisation out of the deep hole that management is intent on driving us deeper into) - even worse, once I've GOT the good stuff in position I then have to deal with mangelement who recall the wheels falling off and don't trust anything/keep interfering with plans to keep the good stuff running

They then get offended when I get sweary and threaten bodily harm. I fully understand and sympathise with Simon's stories and just wish that it would be possible to arrange a few open elevator shafts

And of course - if you get everything running well, your reward is to get your budget cut whilst the best way to have oodles of money thrown at you is to have a "turd bomb" explode under the the director's chair

[juice]

Re: '"how far can I push this before it gets silly". A true engineer'

> He literally spends several minutes talking about "while you can do this, there's really no reason other than curiosity to do so" adding "the processor is a huge bottleneck and is unable to deliver at full speeds. He goes on to test it in several configurations that highlight this fact.

So, what we're saying is that the guy knew that this wouldn't work, and spent $5000 to show that yep, as per the documented hardware specs and all the existing empirical evidence from nearly a decade of people tinkering with Pi hardware, this wouldn't work.

And he then spent several minutes in his video talking about this, to pad out the runtime.

Between that, and the deliberately OTT cover photo for the video, I don't think he's trying to see "how far I can push this before it gets silly". Instead, I think he's producing deliberate clickbait videos to drive traffic to him and maximise the amount of money he gets from sponsorship and referral links.

And, y'know, good luck to him with that. But my time, money and general interest will go into watching videos which show something more interesting!

(Such as "How to drink tea in a tank". Which is sponsored by World of Tanks, oddly enough. https://www.youtube.com/watch?v=cyGVR95P8t0)

[werdsmith]

Re: '"how far can I push this before it gets silly". A true engineer'

Yes @juice. That’s how youtubers operate and have done since YouTube.

And he hasn’t spent any of his own money, just part of a marketing budget and given good value for it.

[Cederic]

Re: '"how far can I push this before it gets silly". A true engineer'

Think of it as educational entertainment. People will watch it out of interest and learn a little along the way.

Meanwhile he's covered his costs and given people something to talk about that isn't politics, religion or curry.

[StargateSg7]

Re: '"how far can I push this before it gets silly". A true engineer'

The whole problem is that Raspberry PI's suck as computing platforms when you can simply buy an AMD 6-core 4650g/u series laptop APU (CPU + GPU) for $220 USD AND LESS and it waaaaaaaay outperforms the PI!

Li-Ion Batteries are cheap these days and a PC104-type board powered just from small form factor Li-Ion batteries can EASILY hold EVERYTHING you need for high performance engineering of IOT (Internet of Thngs) and robotics systems. I like to use Canon Camera batteries! They are small and you can get them EVERYWHERE and at their power levels they can easily power a PC-104 board for hours to days at a time!

V

[Roland6]

Re: '"how far can I push this before it gets silly". A true engineer'

>But in this instance, anyone with any Pi experience would be able to tell you that the throughput is far too small for that type of use. You don't have to build it to know that.

Well he did step back and observe that the performance was adequate for the 1Gbps network - which means a decent/good enough product based on the Pi could be built for the home market...

As for cost, well he did choose some large and relatively expensive storage, swap these for a few smaller and cheaper units...

It would be interesting to see how this compares with the NAS units sold in PC World etc.

[Blackjack]

Re: '"how far can I push this before it gets silly". A true engineer'

I think you are confusing Wozniak with Steve Jobs who made the Mac III such a disaster because he remove the fan because it was too noisy. Wozniak actually made things that worked, it was Steve who just wanted things to be "cool."

[Flightmode]

Re: '"how far can I push this before it gets silly". A true engineer'

If he wanted it to be cool he shouldn't have removed the fan.

[Brad16800]

Re: '"how far can I push this before it gets silly". A true engineer'

Upvote on that Mike, I do find I only read the reg for the comments :)

[Mike 137]

Re: '"how far can I push this before it gets silly". A true engineer'

"I think you are confusing Wozniak with Steve Jobs

Actually I'm not. Woz was an innovator, Jobs was a marketing wizard.

The design of the Apple II disk interface is an example of what I am referring to. It was cunning in that it avoided the use of an expensive chip, but it was not optimum as it occupied the entire CPU capacity during disk transfers by performing in software all the low level operations that a floppy disk interface chip would have done silently without burdening the CPU. Consequently it was great innovation but not brilliant engineering.

[Doctor Syntax]

Re: '"how far can I push this before it gets silly". A true engineer'

Whether it was good engineering depends on whether it was well optimised for its target. It would appear that the optimisation was for cost, not CPU use.

[Blackjack]

Re: '"how far can I push this before it gets silly". A true engineer'

It was one of the first massive popular "portable" computers. Not portable as we think of it today but portable as in you can put it on a box and take it to use it somewhere else. They needed something not hugely expensive that was useful. Steve was the one that came with the requirements on how the machine should be and Wozniak did the best he could.

Something more efficient would have been more expensive.

It says something the Apple ][ was so good, people keep using them for way too long and that combined with the Apple III disaster put Apple in trouble. Fortunately there was still an increasing need for more powerful hardware.

[Anonymous Coward]

Re: '"how far can I push this before it gets silly". A true engineer'

"Consequently it was great innovation but not brilliant engineering."

Nah, it was great engineering.

Somebody somewhere made a statement about the challenges of engineering along the lines of "anyone can build a bridge that doesn't fall down, the hard part of engineering is designing a bridge that doesn't fall down while still coming in at a reasonable cost, being built in a reasonable time, and meeting any number of other trade-offs".

For the Apple II, would it have been nice for the CPU to do other things while disk transfers were going on? Sure. On the other hand, if the resulting product was too expensive for you to afford, then you had no CPU capacity *ever*, because you had no computer.

Which was the correct choice? I don't know, but I do know that I owned a Commodore 64 around that era. Commodore took the "put the smarts in the drive" approach. The C64 drive was like 3x the size and 5x the weight of the Apple II drive. We didn't buy the floppy drive for quite a while due to the cost (got by with the tape drive.. ugh!). I saw plenty of Apple II drives, but very few C64 drives.

[Our Lord and Savior Rahl]

Re: '"how far can I push this before it gets silly". A true engineer'

"A space engineer is a professional practitioner who uses scientific knowledge, mathematics, physics, astronomy, propulsion technology, materials science, structural analysis, manufacturing and ingenuity to solve practical problems in space."

-Wikipedia 2077

[yetanotheraoc]

Re: Coincidence

"I found his blog last night ..."

Not a coincidence. His SEO is working perfectly.

[HereIAmJH]

Re: Coincidence

it became apparent quite quickly he wasn't so much interested in solutions, more "how far can I push this before it gets silly". A true engineer.

Not an engineer, a YouTuber. Taking sponsorship money to build over the top projects for entertainment. Not that entertainment is bad.

It's not like any of the Pi limitations that he found were unknown. And SSDs for speed? Can you even do 10g NICs or link aggregation with a Pi? Fast access on a NAS does no good if you can't serve it.

[Alan Brown]

Re: Coincidence

"And SSDs for speed?"

No. SSDs for low power consumption

8TB spinners draw 8-15W depending on speed and even more when randomly seeking.

8TB SSDs draw under 1 watt on average (2-3 when writing)

[Probie]

Re: Coincidence

Depends on the ssd’s you buy. The ones below take just under 6 watt (and are enterprise grade)

https://www.kingston.com/unitedkingdom/en/ssd/dc450-data-center-solid-state-drive?capacity=1.92tb

It’s important you get that right. One works with a raspberry pi on a usb slot, the other does not, as I know the hard way. :(

[pica]

Re: Coincidence

I have made a lights-out controller using a pi model B+ and me brother that made a mount for it so the RJ45 port was externally accessible via an unusable card slot and a USB car with internal ports (proper ports not a pins). Used one for power and one to wake on usb/ip over usb....not the most elegant solution, but does work.

[LateAgain]

Re: Coincidence

You should probably start by looking at the PiKVM project.

[Pascal Monett]

Even if money is not an issue

A Synology DS420+ goes for around €350. Add 4 8TB SSDs at €830 a pop and you're still €1300 better off and you have a product for which bottlenecks will not occur so often.

I understand what a pet project is, but sometimes you need to look at the cold, hard numbers first.

[A Non e-mouse]

Re: Even if money is not an issue

In my experience, I see a lot of fancy/complex IT where a geek said "Oh, look how clever I am" without bothering to ask the fundatmental question of "Should I do this?"

[heyrick]

Re: Even if money is not an issue

Alternatively, "should I not so this?" is a question to ask.

Sometimes the only reason to not do something is common sense and/or logic. But living by those rules alone makes life boring. Where would we be without people crazy enough to think of controlling a massive explosion and turning it into a way to get into orbit... or to heat up water to spin a turbine to generate electricity?

[werdsmith]

Re: Even if money is not an issue

Should I do this?

Well, if you are Jeff then of course you should, because he is in the business of getting his YouTube videos seen and stuff like this gets attention. He's not out to put a storage server on the market.

And the rest of us who are interested can vicariously enjoy.

[Anonymous Coward]

Re: Even if money is not an issue

And anyone who vaguely thought it might be a good idea can realise such things need a bit more thought, before wasting their own money!

[werdsmith]

Re: Even if money is not an issue

I think anybody interested in watching already knew that. That’s not the point of doing this stuff.

[Stoneshop]

Re: Even if money is not an issue

Or that you just need a large, robust, portable NAS, with speed lower on its list of properties.

There are drive bays that take 6 2.5" drives in a single 5.25" drive slot, which means you can mount such a drive bay in one of those enclosures for an external CD-ROM drive or a 5.25" harddisk, the circuit board with the CM4 behind it connected using short SATA cables. Those enclosures tend to have a power supply built in as well, leaving you to only fit Ethernet and USB connectors to the rear.

And mount a carrying handle, otherwise it's not portable.

[Charlie Clark]

Re: Even if money is not an issue

In this case, there's no need to go that far: bandwidth has always been limited on the Pis, basically by design. Not that it would be that expensive to have a separate controller, but it would bump the BoM noticeably and is totally irrelevant for 99.99% of all Pi projects.

[werdsmith]

Re: Even if money is not an issue

The fact that it’s irrelevant to pi projects is irrelevant to making a YouTube video to attract views.

[Tromos]

Re: Even if money is not an issue

Adding 4 8TB SSDs at 830 a pop will not give you 48TB. It is quite easy to be 1300 better off when you've skimped on two lots of 830.

Sometimes you need to look at the cold, hard numbers CAREFULLY first.

[Mike 137]

"overall, the device's performance was ... mostly disappointing"

Not surprising really. For all its multiplicity of peripherals, the Pi Compute processor's performance is little better than that of my most ancient (2010) palmtop.

[DougMac]

Re: "overall, the device's performance was ... mostly disappointing"

Plus, add on NIC port that can barely do 200Mbps.

The bang for the buck for the Pi is pretty amazing, and it gets by with so little because so many applications need so little.

But storage isn't one of those things.

[druck]

Re: "overall, the device's performance was ... mostly disappointing"

The Pi 4B is showing an iperf result of 968Mbps over gigabit Ethernet.

The Pi 3B+ wasn't true gigabit but could still do about 300Mbs.

[Wellyboot]

Re: "overall, the device's performance was ... mostly disappointing"

Yes the 4B can handle near enough the full gigabit and the compute can probably do a useful percentage of the PCIe bus.

The point here being that the Pi-4 compute can only run the ethernet and PCIe together at a reduced rate before bogging down so don't expect the same performance as specialised NAS controllers costing 10x+ the price.

Having said that, there are two important points to be made

(1) the Pi/PCIe setup used here represents only a small % of the total cost + SSD price per TB will continue to drop.

(2) for home use now, hanging a big spinner from a 4B makes a perfectly adequate NAS backup* solution for under 100GPB plus whatever you can afford in big disks (and for the love of $deity$ don't use shingled disks, they're the cheapest for a reason).

NAS with RAID isn't for backup, it's for capacity, speed & some failure tolerance.

[david 12]

Re: "overall, the device's performance was ... mostly disappointing"

I've got a $250 NAS and a $600 NAS and a $1000 NAS and a $2000 server. And my opinion is, that when you buy a $250 computer, you get a $250 computer.

[m-k]

But overall, the device's performance was ... mostly disappointing.

doesn't matter, what matters is publicity. But you can't blame anybody, cause capitalism, etc., eh?

[Jimmy2Cows]

Re: But overall, the device's performance was ... mostly disappointing.

What's your point?

Are you annoyed that he tried something to see how it would turn out? That it failed? That he's lucky enough to have some sponsorship giving him the fiduciary flexibility to try random stuff? That he blogs wacky ideas to make a buck...?

[Anonymous Coward]

Re: But overall, the device's performance was ... mostly disappointing.

that the media fell for a plan (again) to make news out of nothing to generate interest, thus traffic, thus profit. To me it's no different to showing your hairy (or shaven, whatever) ass on your tube (and I bet the first couple of such created traffic too, until it became meh so they marched on showing something else unseen before, etc. etc).

[werdsmith]

Re: But overall, the device's performance was ... mostly disappointing.

Nobody has fell for anything apart from a few miserable sods getting butt hurt about it.

[BinkyTheMagicPaperclip]

Not surprised

The Pi is useful for lower end, low power server stuff and I/O interfacing.

Recently I've tried using it to replace a Core2Quad Unix desktop with a fanless silent Linux based Pi 4 (sadly BSD on the Pi has unaccelerated Xorg).

I would describe the experience as painful. Raspbian is quite fragile, I see issues with USB that don't appear on an x86 PC, and it's considerably slower than the 2008 era Core2Quad system.

The takeaway so far is 'Windows is far better than this', although to be fair I expect Linux on x64 wouldn't experience all the issues I'm seeing.

Recommendations for silent small x64 boxes welcome.

[werdsmith]

Re: Not surprised

That last thing that Raspberry Pi OS based on Buster (it hasn't been called Raspbian for ages) is, is "fragile". It's as strong as it needs to be for an educational/experimental system, and that's pretty strong.

The new Bullseye based version needs to mature a bit.

I get along with it just fine for desktop use.

[J. Cook]

Re: Not surprised

I've run into problems with the Pi 4 bogging down trying to play a streamed video at 4K; If I drop the resolution down to 1080P? trucks right along.

I will state that a Pi4 is not something that's good as running a plex server, though, especially if the thing has to perform any sort of transcoding of the video source. :(

[doublelayer]

Re: Not surprised

"It's as strong as it needs to be for an educational/experimental system, and that's pretty strong."

This depends heavily on the goals you have in mind. If, as the original post suggests, it is meant to represent the power that Linux brings, there are several places where it really isn't there, especially for a nontechnical user. For example, put any nontechnical user in front of a Pi running a desktop environment and they'll quickly notice a few problems. For example, they might try to watch a video online using Firefox, something the cheapest of machines can do easily, and the Pi won't handle it. We as technical users familiar with the Pi understand that Firefox doesn't have hardware video acceleration support enabled by default, it doesn't work that well when turned on anyway, and that, if you want to use it, you have to use Chromium. The average user doesn't already know this and might well ask why that problem hasn't been fixed.

The Raspberry Pi is a great machine, and in comparison to the original comment's old processor, it is much more efficient. It must however be acknowledged that while what it does is more efficient, it also does less. Someone who used the resources of a desktop processor may find that the Pi's IoT-class SoC is not sufficient for their needs. Nobody expects to make an urgent cross-continent trip on a bicycle, and nobody expects to use a Pi for something processing-heavy.

[werdsmith]

Re: Not surprised

“Goals in mind”

As I stated, educational/experimental system.

[doublelayer]

Re: Not surprised

Educational/experimental could mean a lot of different things. For example, educational system could mean the Pi foundation's original dream of a desktop used in a school for ease of maintenance. The original Pi was almost entirely incapable of that, and the newer Pi, while it can do most of the tasks you might expect of a school computer, isn't always capable (running some bloated online learning platform which is a memory hog on a 4B 1GB, for instance). Experimental system is even more vague, as you can experiment with a lot of different things. Experimentation with administering a system, using a lot of server-style software, etc.: a Pi is great for that. Experimenting with machine learning where the instructions recommend a minimum spec for the GPU: think again. Therefore, it still depends on the use case, and a quite detailed one.

[werdsmith]

Re: Not surprised

Experimenting with ML where you can plug a Coral device into a USB port and get decent performance. Think again yourself. There are plenty of ML things you can do with PyTorch and Tensorflow etc even without the Coral.

Anyway, there is plenty of education/experimenting you can do aside from ML. It's not vague at all, it's absolutely true. Nobody expects it to do everything, it's £35 FFS.

The original dream wasn't for a low maintenance desktop in schools. That's BS. The idea is for a device where you can switch out an SD card. It is totally successful at that.

Anyway, the massive global success of the Pi and its supporting backup resources speaks for itself. If you are inherently negative then I can see that you might miss that.

[doublelayer]

Re: Not surprised

I have lots of Pis here, recommend them to others, and think they're great. I don't think I hid that fact, and I'm not sure it makes me "inherently negative". I was simply making the same point you did: "Nobody expects it to do everything, it's £35 FFS."

While you object to my pointing out the negatives, I don't think they're incorrect. You've pointed out that you can plug in something else to do your ML experimentation faster, which you can also do to anything else. And you can also use the Pi to control an external server running your real computation, but the end result is that you are using some other device to do the heavy lifting. Nor did I say that you couldn't do any ML; something small can definitely be run. However, the larger models one might want to experiment with do often require more processing than the Pi can perform in a reasonable length of time. This isn't a problem; the Pi wasn't designed for lots of raw computation. I think that's important to acknowledge when people point out they have difficulties because it doesn't have enough performance.

"The original dream wasn't for a low maintenance desktop in schools. That's BS. The idea is for a device where you can switch out an SD card. It is totally successful at that."

Yes, switching out the SD card, thus easier to maintain. And they really did intend them for schools. If you look at the projects the Pi foundation was excited about in 2012, you can see the kind of thing they had in mind, and they also stated it outright. They didn't create the Pi to run digital signage, nor really for us to use it as a cheap platform for automation. Their goal was use in education, with others' benefits being a bonus. That's still their goal, and the Pi 4 is a lot more capable of the task than a Pi 1B was. Just not every educational use you could consider.

[Charlie Clark]

Re: Not surprised

Do any serious digging in a Pi and you will hit problems. I've been running Kodi with a Pi 1, 2 and now 3 and still have problems with audio dropouts at the start of some shows.

Doesn't mean it's a poor system – it's a game changer – it's excellent for small systems such as a firewall or for prototyping.

[werdsmith]

Re: Not surprised

Yes, it’s not a media streaming platform. Get an NVidia shield if that’s what you want.

[smot]

Re: Not surprised

Running Kodi here on a Pi4 with a Hauppage twin tuner and it's spot on. It's currently storing around 300Gb on a 1Tb 2.5" HD.

It's been rebooted about 4 times tthis year and I stream it onto my lappie while the rest of the family watch Neighbours and Eastenders.

[FatGerman]

Re: Not surprised

Running Kodi here on a 4B. Playes Netflix at 1080p with 5.1 Dolby Surround through a USB sound card. No dropouts, no stuttering. You need a fan though :)

[Rahbut]

It was done for a bit of fun - I don't see the problem here...

[cyberdemon]

YouTubers

It was done so that he could put on that surprised YouTuber face and make yet another pointless clickbaity video to add to the sewer of shite that is YouTube. Why is this being picked up by the reg?

[werdsmith]

Re: YouTubers

Yes, and ?

This is what he intended. It’s a laugh, not something to wet yourself about.

[Aristotles slow and dimwitted horse]

Such opportunity. The like of which you've never seen...

I'm going to set up an, erm... "merch" store for my comments on El Reg and then hawk the url when I post comments. Then I'm going to set up a Patreon page so I can try to convince everyone else to pay for my lifestyle. If that doesn't work, then I guess I'll resort to thumbnails of girls in bikinis to try to lure people in. That seems to be the stock clickbait method on YT when a channel needs to up its stats.

When I bought my project sailboat last year I did actually think of setting up a YT channel to document the renovation. Then I remembered (A) how much I dislike Google, (B) how shit and desperate most of the YT stuff is, (C) that I couldn't actually be arsed and didn't want to be a slave to a camera, and (D) that I don't actually need the money (not that I'd have made much TBH).

[Yet Another Anonymous coward]

Re: Such opportunity. The like of which you've never seen...

Your sailboat reno video needs to include lots of pictures of you in a bikini (recommending you shave your back)

[Anonymous Coward]

Re: Such opportunity. The like of which you've never seen...

It is a shame how so many of the young ladies on Youtube converting and living in a sprinter van don't seem to be able to afford many clothes. Although they do seem to have fancier kitchen surfaces than my house.

[Doctor Syntax]

However a Pi with a single disk and Nextcloud is quite a handy domestic server although over the last few releases log-on seems to have got rather slow - or is it just mine? Anyway there's no need to log on very often.

[hammarbtyp]

The engineering gospel

He who has not seen a product and said "I can build a better one and cheaper" and then spent loads more than the original product proving themselves wrong, may cast the first stone.

[KarMann]

Re: The engineering gospel

He who has not seen a product and said "I can build a better one and cheaper" and then spent loads more than the original product proving themselves wrong, may build the trebuchet to cast the first stone.

FTFY

[Yet Another Anonymous coward]

Re: The engineering gospel

Is there any field of engineering that hasn't dreamed of building a Trebuchet?

It might even make control systems bearable

[Doctor Syntax]

Re: The engineering gospel

An excellent method of removing stray PHBs and beancounters from the vicinity. Pity about the rate of fire.

[Mark #255]

Re: The engineering gospel

Obviously, the engineering challenge is to make the thing large enough to accommodate all the unwanteds in one go.

Never mind the latency, look at the throughput.

[Alan Brown]

Re: The engineering gospel

If it can throw sufficently large quantity of flaming gelatanous oil (napalm)into their midst, the rate of fire isn't quite as important as delivery accuracy and not having it backfire

[Montreal Sean]

Re: The engineering gospel

Don't forget to put a "Built, not bought" sticker on the trebuchet.

[Roland6]

Re: The engineering gospel

Taking care to cover (or remove) the "Made in China" sticker...

[elsergiovolador]

Re: The engineering gospel

Product development will cost more than the product you are trying to copy, but once you actually develop it, so it meets your requirements and is actually repeatable, then you have a chance at making it cheaper in volume.

[hammarbtyp]

Re: The engineering gospel

"then you have a chance at making it cheaper in volume."

Have you been on Kickstarter recently?

[b0llchit]

Speed compensation

He should have added a whole lot of flashy LED strings, doing blinkenlights things without any reason. That would have improved the system's performance. Not in terms of raw speed, but entertainment value while waiting for your data.

And, if all fails, like missing all data, then you can make it shine in a fine red color. See, "performance" is in the eye of the beholder.

[Yet Another Anonymous coward]

Re: Speed compensation

Alternatively he could return the drives to Amazon for a refund while earning $$$ through YT ads and referable links while getting lots of free publicity from el'reg

[geerlingguy]

Re: Speed compensation

I can guarantee I do not do scummy things like buy a product, use it for a review/video then return it.

In this case, I'm going to be building a more serious storage server with combined SSDs/HDDs for some expansion next year, as my current NAS is over 70% full.

[HereIAmJH]

Re: Speed compensation

What if he used the LEDs to do the data transfer. Fiber-less fiber networking.

[Version 1.0]

Cloud storage is being upgraded

Google Backup and Sync and Microsoft OneDrive are both being upgraded to stop them working on Windows 7 so we'll probably see a lot of people chasing their own cloud storage so that they can keep using it because it doesn't get "upgraded" to stop it working any more.

[Yet Another Anonymous coward]

Re: Cloud storage is being upgraded

1, Buy Synology box and stick 4x8tb spinning rust drives in it, go to pub

2, Buy HP microserver tower, install open source build of synology DSM, stick 4x8tb spinning rust drives in it, go to pub an hour later than 1

[Pirate Dave]

Not too bad

A video about building a $5000 storage device that works but is slow, isn't too naff. I've watched AvE tear into brand-new, relatively expensive tools in his BOLTR vids, and thought "there's no way he's getting that thing back together and working". And sometimes he doesn't. But I've never begrudged the $5 I donate every month - it's cheap entertainment.

[Yet Another Anonymous coward]

Re: Not too bad

Keep yer dick in a vice stick on the ice

Skookum

[John Brown (no body)]

Re: Not too bad

"But I've never begrudged the $5 I donate every month"

ICBA to do the conversion rate, but £5.99 is the cheapest Netflix sub in the UK, which is the one I have, and that's more entertaining than YouTube.

[gotes]

Re: Not too bad

Personally I get a lot more entertainment from a single YouTube channel than I do from everything on Netflix.

[Pirate Dave]

Re: Not too bad

OK, fair choice. It's your money, spend it how you want.

For the record, I don't give the money to YouTube, I donate it to AvE via Patreon (and watch his videos there). It gives him the $$$ to buy the $500 drills that he dissects into a pile of components, and the beer he needs to fuel the effort. It's like directly supporting an "artist" as opposed to sending it to a large, faceless entertainment corporation.

[Yet Another Anonymous coward]

Re: Not too bad

That's the challenge. Get AvE recognised as a national treasure and give him a Canadian Arts Council grant, (he will need to do half the video swearing in French - Tabac !)

[werdsmith]

Re: Not too bad

Absolutely nothing on Netflix I am interested in. Seems it represents the ultimate dumbing down of media.

[picturethis]

A real engineer..

would have been able to determine the limitations by just looking at the Pi4B's architecture and not wasting money/time on building this monstrosity.

The "GigE" ethernet and the USB3 share the same PCIe bus. That's death for any kind of performance for use as a NAS - As Network requests and drive accesses will throttle each other - reducing throughput. As it is, I believe the Pi4B's GigE really is around 300 Mb/s at best due to the PCIe bandwidth.

A single SSD works and is the most cost effective, highest performance approach with the Pi's (3b & 4b). Done. Simple. Next.... I have several of these in this configuration and they work great, for what they are/cost.

I could never understand why anyone would put Pis in a cluster - perhaps to learn about multprocessing, distributing workloads, but that's LEARNING, not expecting performance.

I'm hoping Pi 5's will have dedicated PCIe lanes for network and USB3 and an open-source blob. That would be cool... And fix the audio, put an amp on there so that it can drive a speaker directly (my wishlist)

[NotJustAStorageDude]

Re: A real engineer..

This.

Had a decent size ssd kicking around and added it to a Pi4, makes a great little home server and gets sustainable 90MB/s for large file copies; add Plex, Pihole.. then add another pi for resilience as a second internal dns.. and cross replicate.. and you’ve got a datacentre in a box that even a 9 year old can set up; which 15 years ago would have cost £1million. I’m really looking forward to what we’ll be able to achieve with £200 of hardware in another 15 years!

[Yet Another Anonymous coward]

Re: A real engineer..

It's not just the hardware it's the software.

You can play with this at home and run ZFS and K8s and Docker and other cool free stuff that is cutting edge in the corporate world. It means you can high high-school BOFHs that have years of experience in the stuff you need to setup a quick CI server.

Back in my day I was unemployable as an undergrad because I only knew SunOS and the rest of the world was MVS or CICS or Netware or something that you couldn't play with if you didn't have access to $$$$ hardware and licenses.

[James Hughes 1]

Re: A real engineer..

Not this. The gigethernet on the Pi4 is not on the PCI bus, so there is no usb3 contention. It's a peripheral on the SOC itself, so easily hits speeds close to the theoretical max.

[cyberdemon]

Re: A real engineer..

And inside the SOC, is the peripheral bus PCIe or something else?

[ICL1900-G3]

Re: A real engineer..

I'm not. I'll be dead by then.

[timrowledge]

Re: A real engineer..

People that claim to know how to measure it have published results showing Pi 4 Ethernet bandwidth exceeding 900Mb/s. The 300ish figure is what the Pi 4b claimed.

I don’t much care; whatever the number is, it works pretty well for my usage. Which, as it happens, is as a Smalltalk engine attached to my iMac. Why run it on a (certainly slower than iMac) Pi? Because I have particular requirements to use ARM machines.

[geerlingguy]

Re: A real engineer..

A few things incorrect here—first, the NIC on the Pi 4 does not attach to the SoC's PCIe lane at all, it's independent, and can easily saturate a gigabit of bandwidth (943 Mbps in my testing at 1500 MTU).

Second, the tests done in the linked blog post / video were done with a Compute Module 4, which doesn't even include a USB 3.0 chip (like the Pi 4 model B), so the PCIe lane is exposed for use with whatever other functionality you want.

I've been able to get over 4 Gbps of network bandwidth through a CM4 using 2.5G and 4x 1Gbps PCI Express interfaces, but the single PCIe lane is still a limiting factor.

The nice thing is, you still have a separate built-in NIC on the CM4, plus built-in USB 2.0 (wish it were 3.0) on the SoC that are unaffected by PCIe limitations.

[Anonymous Coward]

I'm building a home lab supercomputer and wish AMD would sponsor me for some sweet free epycs and MI100's >.>

[Spencer Tomlinson]

Not all is what it seems!

Back in my early days as a Sys-Op operating Wang mainframe controllers and Unix systems for a well known commodities company we used to experience similar bottlenecks. Renting data from the NYSE, Tokyo etc, over satellite at around $1 per second you can imagine data was super expensive in 1987, that data landed on our network as jobs and we were tasked with backing up our jobs to 9 track tape (yes those big massive reels you seen spinning back and forth in films). Problem is even though we were at full capacity the data was coming in to fast for native 80mb tapes (yeah those tapes stored around 80mb native) and bottlenecks. Now these were the days of thicknet AUI and token rings. So if you went to the loo and a backup job finished and you took 10 minutes for a sit down job thats 600 seconds that your satellite at $1 per second is costing. Needless to say the Sys-Ops worked 24/7 - 365. Bottlenecking also cost around about $300 an hour roughly until Compaq QIC cartridges came along and 9 track went obsolete in weeks as the QIC was better , higher capacity and more reliable storage and a damn sight smaller and could be dispatched by courier and bottlenecks went away and had also developed later to auto loading so no need to rush back from the bog. We’re talking about commodities worth millions of Yen and dollars so the cost had been factored in..... $5000 was pretty much the cost per day and yet we haven’t moved on one bit!!!!

[arachnoid2]

Its sad

That many of you have a downer on innovation to such a degree, how do you think new designs are born without real-world trials and tribulations ?

It may be that it failed to deliver but a bit off out-of-the-box thinking may or may not remove the road blocks, but you will never know if you continue to say "well that cant be done" simply by looking at the specifications. Where do you think the design for the backplane came from it didnt evolve on its own, its inspiration comes from others tinkering on these projects.

Yes he has an income stream to pay for his designs but who in this real world would not want their pet project which you enjoy doing ,paid for by someone else.

[K-Tat]

Capitalism - cute comment.

The Register with all its' viral marketing, internet ads etc... drops the comment "you know - capitalism" as some sort of dis.

You so trendy

[Excellentsword]

Re: Capitalism - cute comment.

¯\_(ツ)_/¯

[Alex Stuart]

"You can view his video below."

*sees that ridiculous 'Youtuber Face' preview*

Hard pass.

[fredesmite2]

raspberry are only meant

for attaching a camera.cam to your dog .

they are $50 toys

[jason_derp]

Complete vs. Good

"Radxa Taco is a complete NAS/Router solution designed based on the Raspberry Pi Compute Module 4."

Proof that a complete solution isn't necessarily a good solution. Kind of like finding "complete assholes" in the wild.

[Anonymous Coward]

So what about the cost?

To everyone ranting about the cost/waste of money... The vast majority of the cost of this 'venture' is the SSD storage. It's not like he won't be able to reuse these drives on a different project/NAS once he's finished, or gets bored, and dismantles his Pi-based setup.

[Henry Wertz 1]

PCIe speed

I wondered just how over saturated that bus was. Well, the Pi CM4 has a single 1x PCIe Gen 2... that is 500MB/sec. That's pretty speedy but yeah, having a whole set of 2GB/sec+ SSDs loaded in there is definitely going to make the bus the limitation 8-)