Kubernetes Cluster Goes Mobile in Pet Carrier

There’s been a bit of a virtualization revolution going on for the last decade or so, where tools like Docker and LXC have made it possible to quickly deploy server applications without worrying much about dependency issues. Of course as these tools got adopted we needed more tools to scale them easily. Enter Kubernetes, a container orchestration platform that normally herds fleets of microservices in sprawling cloud architectures, but it turns out it’s perfectly happy running on a tiny computer stuffed in a cat carrier.

This was a build for the recent Kubecon in Atlanta, and the project’s creator [Justin] wanted it to have an AI angle to it since the core compute in the backpack is an NVIDIA DGX Spark. When someone scans the QR code, the backpack takes a picture and then runs it through a two-node cluster on the Spark running a local AI model that stylizes the picture and sends it back to the user. Only the AI workload runs on the Spark; [Justin] also is using a LattePanda to handle most of everything else rather than host everything on the Spark.

To get power for the mobile cluster [Justin] is using a small power bank, and with that it gets around three hours of use before it needs to be recharged. Originally it was planned to work on the WiFi at the conference as well but this was unreliable and he switched to using a USB tether to his phone. It was a big hit with the conference goers though, with people using it around every ten minutes while he had it on his back. Of course you don’t need a fancy NVIDIA product to run a portable kubernetes cluster. You can always use a few old phones to run one as well.

youtube.com/embed/XudewmfourQ?…

hackaday.com/2025/11/18/kubern…

Cloudflare va giù nel magnifico Cloud! incidente globale in fase di risoluzione

18 novembre 2025 – Dopo ore di malfunzionamenti diffusi, l’incidente che ha colpito la rete globale di Cloudflare sembra finalmente vicino alla risoluzione. L’azienda ha comunicato di aver implementato una correzione e di essere ora nella fase di monitoraggio attivo, dopo una giornata caratterizzata da disservizi, errori intermittenti e problemi sui servizi applicativi e di sicurezza.

L’incidente, iniziato alle 11:48 UTC, ha coinvolto varie componenti dell’infrastruttura Cloudflare, generando rallentamenti, timeout e blocchi a livello globale, con impatti anche su CDN, API, autenticazione e dashboard di gestione.

Di seguito la ricostruzione completa della giornata.

Timeline dell’incidente Cloudflare

11:48 UTC – Inizio dell’incidente
Cloudflare segnala un degrado interno del servizio. Diversi servizi risultano instabili a livello globale e viene avviata l’analisi del problema.

12:03 UTC – Indagine in corso
Cloudflare conferma che alcuni servizi continuano a essere colpiti da errori intermittenti.

12:21 UTC – Prime evidenze di ripristino parziale
Cloudflare osserva miglioramenti, ma i clienti sperimentano ancora errori più elevati del normale.

12:37 UTC – Indagine ancora attiva
Persistono anomalie diffuse sulla rete.

12:53 UTC – Indagine ancora in corso
I tecnici continuano a lavorare senza individuare ancora una soluzione definitiva.

13:04 UTC – Disattivato temporaneamente l’accesso WARP a Londra
Durante i tentativi di mitigazione, Cloudflare disattiva WARP nella region di Londra. Gli utenti locali non riescono a connettersi tramite il servizio.

13:09 UTC – Problema identificato
Cloudflare comunica di aver individuato la causa dell’incidente e di aver iniziato a implementare una soluzione.

13:13 UTC – Ripristino parziale di Access e WARP
Cloudflare ripristina Access e WARP, riportando i livelli di errore alla normalità.
WARP Londra torna operativo. Si continua a lavorare sui restanti servizi applicativi.

13:35 UTC – Problemi ancora presenti
I servizi applicativi non sono ancora stati ripristinati completamente.

13:58 UTC – Lavori in corso
Continuano gli interventi per riportare online i servizi rimanenti.

14:22 UTC – Ripristino dashboard
La dashboard Cloudflare torna operativa, anche se i servizi applicativi mostrano ancora instabilità.

14:34 UTC – Continuano i lavori di ripristino
Il team tecnico prosegue con le attività necessarie a ripristinare tutti i servizi.

14:42 UTC – Cloudflare: “Incidente risolto, monitoraggio in corso”
Cloudflare annuncia di aver implementato una correzione definitiva. Tutti i servizi dovrebbero tornare progressivamente alla normalità, ma rimane attivo il monitoraggio post-incident.

Impatti dell’incidente

L’incidente ha avuto un impatto rilevante su scala globale. Molti siti web serviti da Cloudflare (tra cui redhotcyber.com) hanno mostrato rallentamenti, pagine non raggiungibili, errori 502, 522 e 526, oltre a problemi di caching e routing.

I servizi di sicurezza e autenticazione, come Cloudflare Access e WARP, sono stati fortemente colpiti nella prima fase. In alcune regioni gli utenti non hanno potuto autenticarsi o accedere alle risorse protette.

Anche Downdetector ha mostrato malfunzionamenti, rendendo difficile monitorare l’ampiezza dell’incidente e contribuendo alla percezione di un blackout molto esteso.

L’infrastruttura globale Cloudflare ha registrato rallentamenti e instabilità, con ripercussioni a catena su servizi non direttamente ospitati sulla piattaforma.

Che cos’è Cloudflare WARP

Cloudflare WARP è un servizio sviluppato da Cloudflare con l’obiettivo di migliorare la sicurezza e le prestazioni della connessione Internet degli utenti. A differenza delle VPN tradizionali, che puntano principalmente a fornire anonimato instradando tutto il traffico attraverso server remoti, WARP è progettato per rendere la navigazione più veloce, stabile e protetta, senza appesantire la connessione.

Come funziona

WARP utilizza il protocollo WireGuard, noto per essere leggero, rapido e altamente sicuro. Il traffico viene instradato attraverso la rete globale di Cloudflare, che funge da “strato protettivo” tra l’utente e il web. Questo permette di:

• criptare le connessioni su reti non sicure (come Wi-Fi pubblici),
• ridurre la latenza grazie alla rete globale Cloudflare,
• filtrare automaticamente traffico malevolo o sospetto,
• proteggere da intercettazioni e attacchi man-in-the-middle.

Non mira all’anonimato totale come una VPN classica, ma si concentra su sicurezza e stabilità della connessione.

La differenza tra WARP e WARP+

Cloudflare offre due versioni del servizio:

• WARP: gratuito, utilizza la rete standard Cloudflare.
• WARP+: a pagamento, instrada parte del traffico tramite la tecnologia Argo Smart Routing, che sceglie dinamicamente il percorso più veloce tra i server Cloudflare, migliorando ulteriormente la velocità.

Situazione attuale

Secondo l’ultimo aggiornamento, la correzione è stata applicata con successo e i servizi stanno tornando alla normalità. Cloudflare è ora nella fase di monitoraggio attivo e nelle prossime ore potrebbero verificarsi residui di instabilità durante l’assestamento della rete

L'articolo Cloudflare va giù nel magnifico Cloud! incidente globale in fase di risoluzione proviene da Red Hot Cyber.

Humane Mousetrap Lets You Know It’s Caught Something

“Build a better mousetrap and the world will beat a path to your door,” so goes the saying, but VHS beat Betamax and the world hasn’t been the same since. In any case, you might not get rich building a better mousetrap, but you can certainly create something more humane than the ol’ spring’n’snap, as [nightcustard] demonstrates.

The concept is the same as many humane mousetraps on the market. The mouse is lured into a confined cavity with the use of bait, and once inside, a door closes to keep the mouse inside without injuring it. [nightcustard] achieved this by building a plastic enclosure with plenty of air holes, which is fitted with a spring-loaded door. When a mouse walks through an infra-red break beam sensor, a Raspberry Pi Pico W triggers a solenoid which releases the door, trapping the mouse inside. This design was chosen over a passive mechanical solution, because [nightcustard] noted that mice in the attic were avoiding other humane traps with obvious mechanical trigger mechanisms.

As a bonus, the wireless connectivity of the Pi Pico W allows the trap to send a notification via email when it has fired. Thus, you can wake up in the morning and check your emails to see if you need to go and release a poor beleaguered mouse back into the wild. This is critical, as otherwise, if you forget to check your humane trap… it stops being humane pretty quickly.

If you’re looking for more inspiration to tackle your mouse problems, we can help. We’ve featured other traps of this type before, too. Meanwhile, if you’ve got your own friendly homebrew solutions to pesky pest problems, don’t hesitate to hit up the tipsline.

hackaday.com/2025/11/18/humane…

Casting Metal Tools With Kitchen Appliances

Perhaps the biggest hurdle to starting a home blacksmithing operating is the forge. There’s really no way around having a forge; somehow the metal has to get hot enough to work. Although we might be imagining huge charcoal- or gas-fired monstrosities, [Shake the Future] has figured out how to use an unmodified, standard microwave oven to get iron hot enough to melt and is using it in his latest video to cast real, working tools with it.

In the past, [Shake the Future] has made a few other things with this setup like an aluminum pencil with a graphite core. This time, though, he’s stepping up the complexity a bit with a working tool. He’s decided to build a miniature bench vice, which uses a screw to move the jaws. He didn’t cast the screw, instead using a standard size screw and nut, but did cast the two other parts of the vice. He first 3D prints the parts in order to make a mold that will withstand the high temperatures of the molten metal. With the mold made he can heat up the iron in the microwave and then pour it, and then with some finish work he has a working tool on his hands.

A microwave isn’t the only kitchen appliance [Shake the Future] has repurposed for his small metalworking shop. He also uses a standard air fryer in order to dry parts quickly. He works almost entirely from the balcony of his apartment so he needs to keep his neighbors in mind while working, and occasionally goes to a nearby parking garage when he has to do something noisy. It’s impressive to see what can be built in such a small space, though. For some of his other work be sure to check out how he makes the crucibles meant for his microwave.

youtube.com/embed/56JHC8Jw8IE?…

hackaday.com/2025/11/18/castin…

Cheap VHF Antenna? Can Do!

The magnetic loop antenna is a familiar sight in radio amateur circles as a means to pack a high performance HF antenna into a small space. It takes the form of a large single-turn coil made into a tuned circuit with a variable capacitor, and it provides the benefits of good directionality and narrow bandwidth at the cost of some scary RF voltages and the need for constant retuning. As [VK3YE] shows us though, magnetic loops are not limited to HF — he’s made a compact VHF magnetic loop using a tin can.

It’s a pretty simple design; a section from the can it cut out and made into a C shape, with a small variable capacitor at the gap. The feed comes in at the bottom, with the feed point about 20 % of the way round the loop for matching. The bandwidth is about 100 MHz starting from the bottom of the FM broadcast band, and he shows us it receiving broadcast, Airband, and 2 meter signals. It can be used for transmitting too and we see it on 2 meter WSPR, but we would have to wonder whether the voltages induced by higher power levels might be a little much for that small capacitor.

He’s at pains to point out that there are many better VHF antennas as this one has no gain to speak of, but we can see a place for it. It’s tiny, if you’re prepared to fiddle with the tuning its high Q gets rid of interference, and its strong side null means it can also reduce unwanted signals on the same frequency. We rather like it, and we hope you will too after watching the video below.

youtube.com/embed/JYIU0Nxn8fg?…

hackaday.com/2025/11/18/a-vhf-…

Congratulations to the 2025 Component Abuse Challenge Winners

For the Component Abuse Challenge, we asked you to do the wrong thing with electrical parts, but nonetheless come out with the right result. It’s probably the most Hackaday challenge we have run in a long time, and you all delivered! The judging was tight, but in the end three projects rose up to the top, and will each be taking home a $150 DigiKey gift certificate, but that doesn’t mean you shouldn’t give all of the projects a look.

So without further ado, let’s check out the winners and all the others that tickled the hacky regions of our judges’ brains.

Prize Winners

[Miroslav Hancar]’s LED Candles was a shoe-in, at least if you watched the video demo. It presents itself as a simple LED on a round PCB with a coin cell, but then when [Miroslav] lights it with an actual lighter, it starts glowing. (And maybe smoking just a little bit.) He makes both single-LED and quad-LED varieties, and they’re both gems.

The component misuse is an old favorite: the diode’s forward voltage drop depends on the temperature, and if you measure the voltage across the current-limiting resistor, you can read this voltage and determine when someone is setting fire to your LED. A bonus of the single-LED configuration is that if you touch the LED’s leads, your finger shunts some of the current, and you can “snuff” the LEDs out. And while we’ve seen similar LED hacks before, the addition of actual fire to this one seems to have warmed our judges’ hearts.

[Luke J. Barker]’s Need an Electrical Slip Ring? is simplicity itself, and appears to have been born of the mother of invention. [Luke] was making a VertiBird helicopter toy, which spins around on the desk and rises and falls with joystick control. And for that, he needed a slip ring. Enter the humble audio jack, which fills the job nicely, transmitting power to the rotating helicopter without twisting up wires in the process. There’s not much magic here, but it’s a fantastic idea when you need something to spin.

On the other end of the spectrum, [Craig D]’s Boosting voltage with a cable looks like it shouldn’t work at first, but it does. In most step-up-voltage setups, you’re storing the energy in either a capacitor or inductor, and switching it in and out of the circuit to hop the voltage up. Here, the energy isn’t ever really “stored” as much as it’s “in flight”.

A circuit sends a pulse down a long length of coaxial cable that is left open at the other end. The pulse reflects off the open end and heads back toward the voltage driver, which then fires off another pulse at just the right time to make the travelling wave a little bit bigger, and this continues. It’s like pushing a swing – adding a little extra oomph at just the right time can build up. There’s a lot of cool physics here, a nice simulation that actually ends up corresponding very well with reality, and in the end the pulse timing isn’t rocket science, but rather figuring out the resonant frequency along the coax. And it works well enough to light up two neon bulbs in series (~140 V) off of a 15 V power supply.

Honorable Mentions

We got way more cool entries than we have prizes, so we try to round them up into categories and give them a little time in the sun.

Out of Spec

Normally, a 555 timer oscillator circuit relies on filling up a capacitor with a current that’s throttled through a resistor. How can you make it go faster? Make the capacitor smaller and the resistor less resistive. What happens when you get rid of them both entirely, relying on stray capacitance and the resistance of whatever wire you’re using? That’s what [MagicWolfi] aimed to find out with his Ludicrous 555 project.

The IC-Abusing Diode Tester is an absolutely horrible circuit. Nothing in it works like it should. The only reason the IC doesn’t burn up is that it’s more robust than the datasheet promises, and the battery used has such a high internal resistance that it can’t source that much current anyway. Parts are powered by leakage current, and below their minimum voltage. [Joseph Eoff] counts seven values that are out of spec in this single historical circuit, so that’s gotta count for something.

Junk Box Substitutions

You need GPIO lines, but you have a UART. [Ken Yap] proposes repurposing the DSR, DTR, RTS, and CTS lines as inputs and outputs, and he writes code to make them do his bidding.

Or maybe you’re working on self-assembling robots and you need some magic glue to hold different modules together. [Miana]’s Low-melt-solder connected robots is half research project, and half hack. Resistors are used to melt solder, magnets align the parts together, and when it all cools down, it’s as if two modules are brazed together. This one’s a lot more than a hack, but we’re honored to have it entered in the contest anyway!

Bizarro World

We honestly thought we’d get more entries that made use of the duality of most sensors / emitters. Instead, we got two. [Nick]’s Better Than Bluetooth does the LED-as-photosensor trick, and concludes that it’s better than Bluetooth if expense, limited range, and frustration are what you’re looking for in a data link. Meanwhile, [Kauz] proves that electromagnets are also pickups by building a guitar pickup out of six relays.

Side Effects

Everything is a fuse if you run enough current through it. [JohnsonFarms] pushed conductive filament to the melting point, and calibrated it along the way. While roasting a hotdog with mains voltage and a couple forks, [Ian Dunn] discovered that if you stick some LEDs in it, they light up.

Most old op-amps oscillate out of control when given feedback, unless you damp it down a bit with a capacitor. [Adrian Freed] found an op-amp lousy enough that would do this at audio frequencies, and used it to reimagine a classic noisemaker.

Finally, while you should probably avoid the metastable middle-zones between digital one and zero, [SHAOS] combines unbuffered NAND and NOR gates to tease out a third logic state. [Bob Widlar] would be proud.

Thanks All!

As always, we had more great entries that we could feature here, so head on over to Hackaday.io and check them out. And thanks again to DigiKey for providing our top three with $150 gift certificates. If you’re looking for your chance to show off a project that you’re working on, hang on for a while because we’ll be starting up a new contest in early 2026.

A Quick Primer On TinkerCAD’s New Features

TinkerCAD had its first release all the way back in 2011 and it has come a long way since then. The latest release has introduced a raft of new, interesting features, and [HL ModTech] has been nice enough to sum them up in a recent video.

He starts out by explaining some of the basics before quickly jumping into the new gear. There are two headline features: intersect groups and smooth curves. Where the old union group tool simply merged two pieces of geometry, intersect group allows you to create a shape only featuring the geometry where two individual blocks intersect. It’s a neat addition that allows the creation of complex geometry more quickly. [HL ModTech] demonstrates it with a sphere and a pyramid and his enthusiasm is contagious.

As for smooth curves, it’s an addition to the existing straight line and Bézier curve sketch tools. If you’ve ever struggled making decent curves with Bézier techniques, you might appreciate the ease of working with the smooth curve tool, which avoids any nasty jagged points as a matter of course.

While it’s been gaining new features at an impressive rate, ultimately TinkerCAD is still a pretty basic tool — it’s not the sort of thing you’d expect to see in the aerospace world or anything. ut it’s a great way to start whipping up custom stuff on your 3D printer.

youtube.com/embed/nMDNAysi5EE?…

hackaday.com/2025/11/18/a-quic…

In Praise of Plasma TVs

I’m sitting in front of an old Sayno Plasma TV as I write this on my media PC. It’s not a productivity machine, by any means, but the screen has the resolution to do it so I started this document to prove a point. That point? Plasma TVs are awesome.

Always the Bridesmaid, Never the Bride

An Egyptian god might see pixels on an 8K panel, but we puny mortals won’t. Image “Horus Eye 2” by [Jeff Dahl]The full-colour plasma screens that were used as TVs in the 2000s are an awkward technological cul-de-sac. Everyone knows and loves CRTs for the obvious benefits they offer– bright colours, low latency, and scanlines to properly blur pixel art. Modern OLEDs have more resolution than the Eye of Horus, never mind your puny human orbs, and barely sip power compared to their forbearers. Plasma, though? Not old enough to be retro-cool, not new enough to be high-tech, plasma displays are sadly forgotten.

It’s funny, because I firmly believe that without plasma displays, CRTs would have never gone away. Perhaps for that I should hate them, but it’s for the very reasons that Plasma won out over HD-CRTs in the market place that I love them.

What You Get When You Get a Plasma TV

I didn’t used to love Plasma TVs. Until a few years ago, I thought of them like you probably do: clunky, heavy, power-hungry, first-gen flatscreens that were properly consigned to the dustbin of history. Then I bought a house.

The house came with a free TV– a big plasma display in the basement. It was left there for two reasons: it was worthless on the open market and it weighed a tonne. I could take it off the wall by myself, but I could feel the ghost of OSHA past frowning at me when I did. Hauling it up the stairs? Yeah, I’d need a buddy for that… and it was 2020. By the time I was organizing the basement, we’d just gone into lockdown, and buddies were hard to come by. So I put it back on the wall, plugged in my laptop, and turned it on.

I was gobsmacked. It looked exactly like a CRT– a giant, totally flat CRT in glorious 1080p. When I stepped to the side, it struck me again: like a CRT, the viewing angle is “yes”.

How it Works

None of this should have come as a surprise, because I know how a Plasma TV works. I’d just forgotten how good they are. See, a Plasma TV really was an attempt to get all that CRT goodness in a flat screen, and the engineers at Fujitsu, and later elsewhere, really pulled it off.

Like CRTs, you’ve got phosphors excited to produce points of light to create an image– and only when excited, so the blacks are as black as they get. The phosphors are chemically different from those in CRTs but they come in similar colours, so colours on old games and cartoons look right in a way they don’t even on my MacBook’s retina display.

Unlike a CRT, there’s no electron beam scanning the screen, and no shadow mask. Instead, the screen is subdivided into individual pixels inside the flat vacuum panel. The pixels are individually addressed and zapped on and off by an electric current. Unlike a CRT or SED, the voltage here isn’t high enough to generate an electron beam to excite the phosphors; instead the gas discharge inside the display emits enough UV light to do the same job.

Each phosphor-filled pixel glows with its own glorious light thanks to the UV from gas discharge in the cell.
Image based on “Plasma-Display-Composition.svg” by [Jari Laamanen].Still, if it feels like a CRT, and that’s because the subpixels are individual blobs of phosphors, excited from behind, and generating their own glorious light.

It’s Not the Same, Though

It’s not a CRT, of course. The biggest difference is that it’s a fixed-pixel display, with all that comes with that. This particular TV has all the ports on the back to make it great for retrogaming, but the NES, or what have you, signal still has to be digitally upscaled to match the resolution. Pixel art goes unblurred by scanlines unless I add it in via emulation, so despite the colour and contrast, it’s not quite the authentic experience.

For some things, like the Atari 2600, the scanline blur really doesn’t matter. Image: “Atari 2600 on my 42 inch plasma TV” by [Jeffisageek] The built-in upscaling doesn’t introduce enough latency for a filthy casual like me to notice, but I’ll never be able to play Duck Hunt on the big screen unless I fake it with a Wii. Apparently some Plasma TVs are awesome for latency on the analog inputs, and others are not much better than an equivalent-era LCD. There’s a reason serious retro gamers pay serious money for big CRTs.

Those big CRTs don’t have to worry about burn in, either, something I have been very careful in the five years I’ve owned this second-hand plasma display to avoid. I can’t remember thinking much about burn-in with CRTs since we retired the amber-phosphor monitor plugged into the Hercules Graphics card on our family’s 286 PC.

The dreaded specter of burn-in is plasma’s Achilles heel – more than the weight and thickness, which were getting much better before LG pulled the plug as the last company to exit this space, or the Energy Star ratings, which weren’t going to catch up to LED-backlit LCDs, but had improved as well. The fear of burn-in made you skip the plasma, especially for console gaming.

This screen is haunted by the ghost of CNN’s old logo. Burning in game graphics was less common but more fun. Ironically, it’s an LCD. Image: “logo of CNN burnt on a screen” by [Nate]Early plasma displays could permanently damage the delicate phosphors in only a handful of hours. That damage burnt the unmoving parts of an image permanently into the phosphors in the form of “ghosting”, and unless you caught it early, it was generally not repairable. The ghosting issue got better over time, but the technology never escaped the stigma, and the problem never entirely went away. If that meant that after a marathon Call-of-Duty session the rest of the family had to stare at your HUD on every movie night, Dad wasn’t going to buy another plasma display.

By the end, the phosphors improved and various tricks like jiggling the image pixel-by-pixel were found to avoid burn-in, and it seems to have worked: there’s absolutely no ghosting on my model, and you can sometimes find late-model Plasma TVs for the low, low cost of “get this thing off my wall and up the stairs” that are equally un-haunted. I may grab another, even if I have to pay for it. It’s a lot easier to hide a spare flatscreen than an extra CRT, another advantage to the plasma TVs, and in no case do phosphors last forever.

But Where’s the Hack?

Is “grab an old flat screen instead of hunting around for an impossible CRT” a hack? Maybe it’s not, but it’s worth considering, though, because Plasma TVs don’t get the love they deserve. (And seriously, you’re not going to find the mythical 43-inch CRT, even if it technically existed. And you’ll never find a tube that could match the 152” monster Panasonic put out to claim the record back in the day.)

In the mean time, I’m going to enjoy the contrast ratio, refresh rate, and the bonus space heater. I’m in Canada, and winter is coming, so it’s hard to get too overworked about waste heat when there’s frost on your windowpanes.

Featured image: “IFA 2010 Internationale Funkausstellung Berlin 124” by [Bin im Garten].

hackaday.com/2025/11/18/in-pra…

Internet Archive Hits One Trillion Web Pages

In case you didn’t hear — on October 22, 2025, the Internet Archive, who host the Wayback Machine at archive.org, celebrated a milestone: one trillion web pages archived, for posterity.

Founded in 1996 by Brewster Kahle the organization and its facilities grew through the late nineties; in 2001 access to their archive was greatly improved by the introduction of the Wayback Machine. From their own website on Oct 21 2009 they explained their mission and purpose:

Most societies place importance on preserving artifacts of their culture and heritage. Without such artifacts, civilization has no memory and no mechanism to learn from its successes and failures. Our culture now produces more and more artifacts in digital form. The Archive’s mission is to help preserve those artifacts and create an Internet library for researchers, historians, and scholars.

We were curious about the Internet Archive technology. Storing a copy (in fact two copies!) of the internet is no mean feat, so we did some digging to find out how it’s done. The best information available is in this article from 2016: 20,000 Hard Drives on a Mission. They keep two copies of every “item”, which are stored in Linux directories. In 2016 they had over 30 petabytes of content and were ingesting at a rate of 13 to 15 terabytes per day, web, and television being the most voluminous.

In 2016 they had around 20,000 individual disk drives, each housed in specialized computers called “datanodes”. The datanodes have 36 data drives plus two operating system drives per machine. Datanodes are organized into racks of 10 machines, having 360 data drives per rack. These racks are interconnected via high-speed Ethernet to form a storage cluster.

Even though content storage tripled over 2012 to 2016, the count of disk drives stayed about the same; this is because of disk drive technology improvements. Datanodes that were once populated with 36 individual 2 terabyte drives are today filled with 8 terabyte drives, moving single node capacity from 72 terabytes (64.8 T formatted) to 288 terabytes (259.2 T formatted) in the same physical space. The evolution of disk density did not happen in a single step, so there are populations of 2, 3, 4, and 8 T drives in the storage clusters.

We will leave you with the visual styling of Hackaday Beta in 2004, and what an early google.com or amazon.com looked like back in the day. Super big shout out to the Internet Archive, thanks for providing such an invaluable service to our community, and congratulations on this excellent achievement.

hackaday.com/2025/11/18/intern…

Exploring The Performance Gains Of Four-Pin MOSFETs

Over on YouTube [DENKI OTAKU] runs us through how a 4-pin MOSFET works and what the extra Kelvin source pin does.

A typical MOSFET might come in a 3-pin TO-247 package, but there are 4-pin variants which include an extra pin for the Kelvin source, also known as source sense. These 4-pin packages are known as TO-247-4. The fourth pin provides an additional source for gate current return which can in turn lessen the effect of parasitic inductance on the gate-source when switching current, particularly at high speed.

In the video [DENKI OTAKU] uses his custom made testing board to investigate the performance characteristics of some 4-pin TO-247-4 MOSFETs versus their 3-pin TO-247 equivalents. Spoiler alert: the TO-247-4 MOSFETs have better performance characteristics. The video takes a close look at the results on the oscilloscope. The downside is that as the switching speed increases the ringing in the V_ds waveform increases, too. If you’re switching to a 4-pin MOSFET from a 3-pin MOSFET in your design you will need to be aware of this V_ds overshoot and make accommodations for it.

If you’d like to go deeper with MOSFET technology check out Introduction To MOSFET Switching Losses and MOSFETs — The Hidden Gate.

youtube.com/embed/gsPN982hmD8?…

hackaday.com/2025/11/17/explor…