I got my enterprise SSDs (1.6TB, 3DWPD, drawback: SAS3) from our Craigslist-equivalent. First I got 2 for 150€ and had a nice chat with the seller, who seems to had a big box of these. A week later he offered me his remaining 3 disks for another 150€.
On eBay it seems you have to make offers.
Some local info: I also noticed on the German eBay getting Chinese cards to be much more expensive than on eBay in the US. The same seller asking 80€ or 100€ for the same LSI 9400. However, finding the article on eBay.com and then using the item ID on eBay.de allowed me to get it for the better price.
That's pure anecdata. We don't even know your workload or configuration.
Contrary anecdata: I just replaced my old SSDs: 2013 64GB 20nm MLC at 19% wear level and a 2018 500GB TLC at 34% wear level.
Not because they failed, but because I had the OS on a 64GB RAID1 and needed more space.
Only optimization was setting "noatime".
But that's still a horrible small N, so even the comined data is essentially meaningless.
btw, I replaced them with a bunch of HGST DC SS200 1.6TB from 2018, two of which have about as much capacity as your 30 disks. The 15nm MLC NAND is rated for 3 DWPD and has a 3% wear level. The dual ported SAS3 interface is overkill for me.
I went for a 5 disk RAID6, and could replace it another 8 times while still keeping some spare change for a visit at a gourmet restaurant.
Thanks for linking that, I missed it when skimming the original link earlier. I find that quite a heart warming story which makes me want to set up something similar, I was particularly tickled by the thermal receipt printers for sending each other messages.
Yesterday I bought some oat-based milk-like at Aldi for 90c/l (regular price). It's labeled "oat drink", so might not substitute milk. The (literal) "almost milk" product is listed online for 1,09€/l. They also had options based on other stuff for a similar price.
First time I noticed them there, but mind I don't go to Aldi that often.
What surprises me is the 24 GB of DDR4 DRAM on a dual channel memory controller? AFAIK there are only 8 GB or 16 GB modules, no 12 GB modules. At least I can only find 12 GB DDR5 modules listed, but not DDR4.
This means: The system likely uses 3x 8 GB modules. As a result, one channel has two modules with 16 GB total, while the other channel has only a single 8 GB module.
Not sure how big this impact is with the given memory access patterns and assuming [mostly] exclusive single-threaded access. It's just something I noted, and could be a source of unexpected artifacts.
I'm not deep into the details of the AMD DRAM controller, but this detail could cause some of your anomalies. If this was an academic paper, the findings would be borderline invalid. You might want to remove the extra module and run the benchmarks again.
At least once the tests become big enough to have some data in both partitions, the bandwidth will start to matter.
I can't run it right now: My AM4 desktop has a broken Linux dual boot since some Microsoft update seems to have nuked something important a few months ago, and my pure Linux Intel machine sits at 1.2 load with services I can't stop.
If I get around to fixing my dual boot machine, I'll try to remember running the benchmark and dropping you some results by mail.
I agree with the overal position. Though I believe optimizing to collect zero fines is a bad measure.
A fine can be a relatively just mechanism to show that actions have consequences. And even the best people will occasionally make honest mistakes, so they will just get a fine instead of being persecuted for minor offences.
If fines degrade to a revenue stream, it's an indication something else is off with the financial structure inside the government. At least around here fines don't go into some official's private accounts, but I can see how they might "help" an underfunded department. Thinking about it this way, maybe we should consider funneling fines into a separate pool of money. Though I am not sure what to do when the fines are used to fix damage caused by the action (e.g. ecological damage). Governing is hard :(
That's true, at least around here [Germany]. Plus the nasty stuff might leak into the environment over the decades, and it's a pain to dispose of.
IIUC, they replace them with plastics since the plastic is seemingly more ecologically friendly and easier to recycle.
Mind concrete sleepers are what's used these days. You'll find wood only in shunting or cargo yards. (Or museums. Or the US - see linked commen by LeonM).
Source: I randomly met someone involved with that project. A proper train enthusiast can probably elaborate here, but I think I remember the core idea correctly. Also this obviously doesn't necessarily hold globally, though I can imagine many track operators face similar challenges.
Mind that you should not use a Pi with SD card for HA. I'm not sure what the official stance is, but: In my peer group failing SD cards were responsible for a vast majority of issues. Causes are probably a mixture of heavy logging and power outages; either from the grid, or user error not shutting down the Pi before disconnecting power.
At our house I run HAOS in a VM (on a beefy server). My wife uses the app on her phone (as do I), and we have a cheap tablet with the app for guests. On the laptops/desktops we also have access to the web UI.
As the article points out, remote access for the phones can be done via the commercial offering or a VPN (as in our case: wireguard on the OpnSense).
Wholly depends on how much stuff you’ve got. I’ve started having issues with the Linux io scheduler when the auto backups got large (1G+), the thing just hung and needed a power cycle.
Also what SD card you have, the durability between different models differ a lot. The SD card I use for my dashcam been working for years (A "SanDisk MAX ENDURANCE" card), some other cheaper SD cards can stop working properly after just months if you're using them for write-heavy stuff.
Anecdotal agree, but I only use it for my lights (about 20 smart bulbs + 5 sensors + 4 remotes) and I never tinker with it. At least 4 years later, that little Raspberry Pi 3 + ZigBee dongle + sdcard are kinda forgotten about.
Interesting; my Raspberry Pi 3B+ + Zwave dongle + SDcard eventually got too slow (to the point that HAOS updates would fail). Switched it to an old ChromeBox and it's been no problems since.
The BT blob (or WiFi blob, or general radio blob) often controls low level aspects of the actual RF device. So you can often alter frequencies, power levels and other parameters. This results in the device acting out of spec.
Now one might say that's the fault of the person doing the modifications/manipulations, but regulations in various countries require the device to prevent these manipulations.
(N.B. I'm not in the exact business, but that's my take away from looking into the topic some time ago)
That was my takeaway as well when I was working with Bluetooth modules a few years back.
There were a few more reasons too, at least at the time. The companies in the space didn't have a culture of open sourcing, and there's probably no perceived commercial upside to releasing code for a chip like that.
I think often it's also a combination of what you described and big companies being super risk-averse. Why risk breaking regulations if you can just say no?
Then again, lots of Wi-Fi-enabled devices support simply changing their region setting and will happily let you use different Wi-Fi bands or increase signal power.
True but those bands are still official wifi bands somewhere.
On software-defined radios you can often use them way out of spec, way more so than using a forbidden channel. But in a totally different band. A good example is the RTL-SDR stick which was designed to be a TV received but can be used as a wideband SDR these days. That's a receive-only device so it's not that critical to regulators. Once you can transmit, it becomes more of a problem.
An example of a more problematic transmission device is the Raspberry Pi PWM pin. That's been used to transmit all sorts of stuff on many bands. Because it was never designed to transmit anything, it causes all kinds of weird harmonics and artifacts. It's a really bad idea to use it for that. Most people just do it under controlled circumstances.
The hardware itself is flexible and relies on software for things like calibration, setting frequencies, and setting power.
> Seems like security by obscurity.
There are legal standards to meet around protecting the system from users modifying things like output power. They don’t literally say that the source code must be private, of course, but keeping it closed source makes it much easier to demonstrate that you were not making it easy for people to exceed the regulatory limits.
If you document the registers for setting output power (for example) then you’re giving the end user a roadmap for changing output power.
I don't know, I don't think there are many open-source third-party firmware implementations for these kinds of chipsets out there, so it seems pretty well obscured.
Exactly my thoughts. To avoid out-of-regulation tweaking coudn't they just allow signed code to run? How can I be sure that this closed for, i.e. surveillance reasons?
One can publish the source code and still prevent misuse afaik.
Many modern chips do in fact use signed firmware yes.
Technically they could publish the source in that case but I think some patents would prevent them from doing this. Radio stuff is riddled with patents. Also most likely the "why would we" reason. There is no benefit for the manufacturer.
Why would patents prevent open source? As in, they're afraid of patent trolls trawling their code for "infringements"?
If it's concerning their own patents, no need to hide the source code. A patent is literally "letters patent" or public description of an invention. Trade secrets on the other hand I could understand.
It's often a condition of licensing those patents. NDA's, they're huge in the radio and embedded world. It only became a bit more open when the Arduino and Raspberry projects came along. And even there it wasn't fully open, especially with the Raspberry because Broadcom is one of the worst offenders.
I'm surprised how much they were able to open source as-is. I think part of that is that the SoCs powering the raspberry are kinda 'old news', definitely not the bleeding edge kind of embedded tech. The first raspberry basically happened because Broadcom had a whole bunch of old chips they weren't able to sell. Only when it took off they started making some actually for the purpose.
SUSE just linked to the git repository's web UI, and remarked there are random 502 errors happening; probably so they're not notified about "broken links".
[edit] From the "<meta name="description">" tag (I noticed the nice slug when I shared the link with a friend):
Screen is the traditional terminal multiplexer software used on Linux and Unix systems. We found a local root exploit in Screen 5.0.0 affecting Arch Linux and NetBSD, as well as a couple of other issues that partly also affect older Screen versions, which are still found in the majority of distributions.
I got my enterprise SSDs (1.6TB, 3DWPD, drawback: SAS3) from our Craigslist-equivalent. First I got 2 for 150€ and had a nice chat with the seller, who seems to had a big box of these. A week later he offered me his remaining 3 disks for another 150€. On eBay it seems you have to make offers.
Some local info: I also noticed on the German eBay getting Chinese cards to be much more expensive than on eBay in the US. The same seller asking 80€ or 100€ for the same LSI 9400. However, finding the article on eBay.com and then using the item ID on eBay.de allowed me to get it for the better price.