We're building a high-end router. Here's the update for July 2024.

[@tomazzaman]

There seems to be some confusion in regard to the operating system (apologies for that, it's on me), so let me clarify: we will not lock down the device in any way, so you will be able to upload whatever you'll want to run, provided it's compatible with the hardware. So if we decide to not use OpenWRT, you'll be able to set it up on your own. Or any other Linux-based OS, for that matter!

[@bigpod]

ill be that insane person that will install linux server distro on it and then not do routing on it but builds and kubernetes :) so if you will need to get rid if hardware my buildfarm could use ARM builders 🤣🤣🤣🤣🤣🤣🤮

[@csabasipos6525]

Well, that said. The mentioned data path acceleration, how exactly that is going to work? As pretty much all the known similar solutions from Qualcomm, MTK or Marvell never made it into the vanilla kernel, thus it does not work on Openwrt. So are you going to upstream the proprietary driver of NXP into the kernel to make this work? How will that affect the linux network stack as most of these solutions actually cutting through it. Otherwise this is not going to be high end at all. And if you use the proprietary vendor SDK to build the firmware to actually utilize the data path acceleration, it will not be Openwrt, it will just look like it due to the gui, as we have seen many "Openwrt" like crap built around vendor SDKs. The fact you can install Openwrt on a device is true for many routers, yet without the proprietary acceleration techniques, the performance is limited to around gigabit speeds, definitely nowhere near 10 gigs. So it would be very useful to hear how all that is going to come together.

[@ksenchy]

I'm keen to see what you build on your own although I am a fan of dd-wrt... hopefully you can send one of your routers to Brainslayer so he can add official support for it. Cheers from Maribor

[@tomazzaman]

@@csabasipos6525 Actually, you can already download OpenWRT build for our very CPU! NXP provides the kernel on their GitHub page and I did manage to run OpenWRT on the NXP development board.

[@mihak]

@tomazzaman: FreeBSD has sample networking kernel modules that can serve as excellent starting point so you see what would it take. One of the most commonly referenced examples is the "if_em" driver, which is for Intel PRO/1000 Gigabit Ethernet adapters. And you'd get help. BTW, I know I am preaching to the choir here, but there is a reason why all decent networking appliances use FreeBSD and not Linux. The reason is pf (packet filter) that has no equivalent in Linux world. Just do the processing math - multiple 10 Gbps streams pushed through 1,000 filtering rules... Without pf your firewall will have some minor choking issues.

[@laceflowerhw]

PLEASE opt for 40mm fans, theyre much more standard. If the fan fails it should be replaceable, and having to mod the case just to fit something like an A4x10 would suck. My experience with industrial fans are that theyre quite loud and run at high rpms - 40mm should in principle be slightly quieter and produce a slightly lower tone compared to 30/35mm thats easier on the ears. Frankly, im surprised you're not going 80mm or 92mm

[@collectorguy3919]

The best fan is no fan, but otherwise it should be an easy to find size.

[@LordSaliss]

Yes, please use a 40mm fan. It will be so much easier to source replacement parts for.

[@asm2750]

I also agree with this, 40mm is a pretty common size that is easy to find replacements.

[@Spiker985Studios]

(Haven't gotten to the point in the video yet) Additionally, make sure the required voltage is easy to find Fan manufacturers like Noctua have both 5V and 12V 40mm variants

[@mhn23visual]

I would really advise against building a custom OS. Impossible to keep up with the contributions from the community and building and maintaining the initial prototypes are easy, but trust me, building all the edge cases out is hard. I’d much much much rather invest the time to pay knowledgable devs temporarily to get opensense or openwrt running with your hardware, contribute the code. You’ll get “free” updates and security patches as opensense gets updated, rather than having to patch it yourself. Then you can build a custom UI ontop of it. I’m sure there are passionate open source developers out there, that you can pay, that are happy to contribute to your project. Be really careful with the decision to build the software in-house and maybe get some consulting/advise on this from pros in the area.

[@tomazzaman]

Thanks for the feedback. This is exactly why I voice my plans and concerns here. So that I don't make any stupid decisions :)

[@JasonFritcher]

@@mhn23visual exactly my thoughts. As someone who’s maintained a custom Linux distro, there is a lot of work that goes into building and maintaining infrastructure to keep up with your upstream packages and handle release engineering in an efficient manner. Im sure you’d rather spend resources doing more interesting things. :) As a BSD fan, I’d love to see the NXP drivers written for OpnSense and contributed back to the community. This would also eventually make it available to pfSense via FreeBSD, which I’m kinda partial to.

[@michalmarszaleck]

Yes, openwrt/... + private drivers and UI.

[@dvatp]

Long time embedded systems engineer here. Don't take this the wrong way, but in my opinion you're looking at the software/hardware problem backwards. This is not custom military or space grade hardware with the software development budget to match. It's a commodity router. OpnSense represents a huge body of work that very few fledgling companies could afford to build from scratch. And even if such a thing were fiscally practical, why would you bother just to reinvent the wheel? I have more specialized networking needs so I'm probably not in your target market but I'd actually consider buying it for friends and family if it ran OpnSense because that's what I know and that's what works. Sorry if this is harsh, but such is business at times. Best of luck to you and the project.

[@rpsmith]

Makes no sense to me. He has no idea of the rabbit hole he is headed down but I bet future investors can figure it out!

[@mt1104uk]

As he explains in the video, OpnSense uses OpenBSD as the base, and as such he cant use the hardware offload features of the ARM cpu platform they have chosen. Now, he did say they provided firmware and drivers for linux, but I didn't catch if they were open source or not, if they are then they could potentially try to port those over to OpenBSD however if they are just binary blobs then that's a no go. Ultimately this comes down to choosing a non x86 platform to build from.

[@yoloninja4798]

There is a third way - fork opensense, and add support for the arm cores. You don't need to reinvent the wheel, but you still have the dedicated hardware for networking.

[@Kas-_-]

@@yoloninja4798even forking vyOS or IPFire would be better than porting FreeBSD-based OPNSense in this case Why? Because the firmware is ready for *Linux* not any BSD. You cannot just run ELF-compiled files in BSD or NT-based systems, and reverse engineering that ain’t that easy as you’ll have to look into the binary file with a hex editor and at least check out each interrupt or system call. Any call utilizing the Linux API/ABI has to be modified to call an equivalent FreeBSD API/ABI function, assuming a 1:1 translation is possible (and we cannot assume that) Using OPNSense just makes zero sense in this specific case..

[@ahmetkipkip]

Good comment, I'm thinking the same. Reinventing wheel.. Contribute community and make openwrt Frontend better. With optimizations you've mentioned.

[@starrynayt]

"We might fail completely but you have my word, it won't be for our lack of trying". I didn't know I needed this today Tomaz, thank you, and looking forward to your journey in building this router!

[@Xploder270]

Programming your own GUI for OpenWRT should be the way forward I think. Best support for hardware offloading in ARM chips and very leading-edge features and versions (nftables, latest kernels). It's such a great foundation.

[@Hyde-Jahf]

I'm really interested in this project, but you're going down a rabbit hole I was hoping you'd avoid. I see all of your reasons for not picking openwrt or opnsense. But now you have the huge hurdles of creating the software, extending it for the features your customers, making a UI both you and customers are happy with, and maintaining it all. All of which adds cost to both sides and complexity to your side. Personally I wish you'd do the hardware, get it awesome, and just support the openwrt. Then, if enough success, work on a custom software stack. I think many of your audience was assuming this would have a 3rd party stack for software and are planning to run one anyway. Professionally (I'm a former Cisco & Linksys TM, not an engineer but did a lot of work between engineers and PMs) ... I applaud your initiative, but am worried at how much you're now trying to get done.

[@Kitagua]

Totally support your statement. The effort of developing an own OS is huge and will most likely delay the router project a lot. @Thomasz: I would prefer to use existing software. What about starting with it (even if not optimized) and developing and maintaining either some kind of optimization "add-on" or an own OS afterwards? You could benefit from router hardware sales and sell the optimized software as an add-on? This would ensure stable revenues during the early stage of development. Maybe start some performance testing with existing software first before deciding how to proceed. Thank you for discussing these things openly with us!

[@bdg88]

@@Hyde-Jahf Agreed. Better to put the effort into making something existing work. VyOS for example.

[@PugnaEnjoyer]

There are also significant security risks. Writing a whole lot of low level networking stuff from scratch is a recipe for security vulnerabilities.

[@tomazzaman]

Thanks for the feedback, and of course, a project like this won't happen overnight and certainly not yet at this point, it's more of a question to where do I see the company (provided we succeed) in 5-7 years?

[@congenio]

@Hyde-Jahf: This! Actually, I was a bit disappointed that the CPU/platform choice did not take this into consideration in the first place. Considering the amount of work that has gone into FreeBSD-based versions of firewalls like pfSense and OpnSense, I would have liked to see an X64-based platform for this to give even more freedom of choice. It will become especially complex if you have to tailor the software to support specialized hardware features like the networking acceleration present in that NXP chips. It is a venture in itself to create a decent router hardware platform that can do more than the usual N100 box from china, but I think you are looking at a 20+ times bigger effort by adding the software part. You will find it needs a whole infrastructure of people willing to test, develop add-on features and such. I do not question your will to do it, but consider that all of those people will have nothing to start with, i.e. they cannot use a basic set that "just works". When I just look at the discussions going on with OpnSense about how one could group firewall rules in the GUI and this is a big problem because of the GUI framework architecture, I cannot imagine that a CLI-only, restricted version would be enough incentive for people to use it. After all, people use platforms like OpnSense because besides the usual firewall and routing features, they can have: - VPNs: OpenVPN, Wireguard, IPsec - and even then, they ask for Tailscale - Traffic introspection in order to be able to protect their children - Suricata or even Zenarmor - Crowdsec - IP Blocklists - GeoIP blocking - VLAN segmentation - Running a Unifi Controller in a jail to control their access points - etc. (this list is potentially endless, just look at the OpnSense plugin section) If they do not seek for this, they choose something simpler, like a Fritzbox. There was a similar approach by Ugreen earlier this year to get into the NAS market. While I think they build great hardware, their attempts at coming up to par with market leaders like Synology and QNAP is a little pathetic with their own Ugos software. However, they made a wise move to support different NAS OSes by choosing an X64 platform, such that current customers have the option to use Unraid, TrueNAS (Scale or Core), Proxmox or OpenMediaVault, just to name a few. I would seriously re-think that decision.

[@soter190]

Awesome work Tomaž!!! It's a bless every time I see there's a new video from you!!

[@tomazzaman]

Thank you!

[@KiceDz]

Add fins on the bottom of the router, like grill, so it can dessipate heat much easier passively. Futhermore, i really hope the price would be affordable. Other manufacturers already have high end routers, with high prices but they have almost all the features working out of the get-go. Great idea, hope you get it up and working. Much love from Makedonija

[@tomazzaman]

Thank you!

[@bloodaid]

The grill should be at the top, or whatever ”top” the orientatipn of the router whould be, because heat rises.

[@dakoriki]

I'm not a developer and I will not pretend to know exactly how OpenWRT works but if its a big "problem" how the UI of OpenWRT, LuCI looks, you could swap it for something else or fork it and change stuff? Because AFAIK LuCI just changes config files thru UCI. And companies like GL.iNet have built pretty UI's for OpenWRT. Tho I do kinda see a linux distro for routers to be an interesting thing as well

[@blunden2]

There is already a Linux distribution for routers. It's called VyOS. 😄 It's pretty awesome, although it lacks a GUI since its an enterprise distro.

[@unicodefox]

LuCI has a lot of features the UIs of much more expensive routers don't get. For example, with LuCI changes aren't applied immediately, so you can make large, conflicting changes and apply them all at once, revert. It will also automatically revert any changes made if you dont access the web UI within 90 seconds.

[@Xploder270]

@@blunden2 aren't Unifi Gateways and Edgerouters basically Vyatta/VyOS with a nice GUI?

[@uis246]

I haven't watched yet, but I know thay LuCI has skins

[@andibiront2316]

Why not a Front End over something proven like VyOS? You get a powerfull CLI based router and firewall, and need only to invest on a UI

[@ArthurWiebe]

Yeah I already use VyOS wherever possible. We need to find a way to make this work.

[@hexlocation]

VyOS doesn't provide stable images for free, which could be a pretty major road block. Though it does seem like a cool system.

[@blunden2]

​@@hexlocation This would clearly be a custom build anyway so that doesn't matter. There will also be VyOS Stream soon, that is free.

[@nazar5549]

yeah commercial use would probably require a license

[@blunden2]

Yes, that would be great! They could even implement the GUI design concept images that were released.

[@cristinelcostachescu9585]

My personal recommendation for the OS is OpenWRT. I would be fine with the stock OS in the first production version, and then a major OS overhaul after some months with custom UI and all. An OS/kernel that does not take advantage of all of a CPU's features is way worse than a lacking UI. Also, do NOT to go DIY unless there are major issues with available solutions. This introduces way too many variables in development - did the hardware fail or did the OS do something unexpected? I'd recommend isolating your points of failure as much as possible, removing any if possible. Looking forward for the next update!

[@tomazzaman]

Thanks for the tips!

[@stephenreaves3205]

"How many components should a router have?" I dunno, like 20? "This includes resistors and capacitors" Oh, I'm going to be way off

[@smilespray]

I was like, 300? 😂

[@monad_tcp]

sprinkle capacitors like salt in a meat you're cooking

[@monad_tcp]

I swear I saw a video like that, but youtube can't find it because its search sucks, all that stupid investment in stupid AI and it can't find videos

[@sjplabs]

When you are done with "The Ultimate Router" project, please build a printer. 😀🙌

[@tomazzaman]

Oh no, I value my mental health way too much for something like that :)

[@thejumper9303]

Just fork WRT, fork it, throw everything you don't want away and replace it. You don't want to start from scratch. Even less people will contribute to your software then to freebsd. You could also use opensense and add hardware support ;-) I am definetly for the opensense version.

[@bdg88]

This thing should just run VyOS. Best router/firewall OS out there. Ubiquiti runs their own Vyatta fork on their Edgerouter line.

[@rpsmith]

And the Edge Router can only manage to do a subset of what pfSense and OPNsene are capable of and they have been developing that for years!

[@jonathanthorpe3444]

Agree! There’s a lot of scope to build a great web ui around this and also let people leverage all the tools that work with VyOS (e.g. Ansible)

[@YehudaKatz1]

VyOS has come a long way recently and will only get better with more people using it. I think you should definitely look at it.

[@efimovv]

OpenWRT UI is just UI - it works via API to the core system. So, you can write your own UI and even make it closed source.

[@sid.h]

This is the answer. Extend/modify OpenWRT - it's open after all - don't recreate it!

[@hex2307]

​@@sid.hI think vyos would be a much better base. It has more advanced features supported e.g. BGP/VRF etc without needing any packages. It already has some support for hardware offloading although I've never properly tested it. It also already has a web api.

[@psylenced]

I won't repeat all the other comments about DIY software as they have it pretty well covered. One thing I'll add however is security. You want your router to be secure. DIY means much more chance of security vulnerabilities or misconfigurations. Unless you have a team ready to handle critical security issues and be able to knowledgeably debug and fix those - it's much better to run pre-existing router software and just run your own GUI.

[@sagarchandarana3608]

How about NixOS? A declarative approach at the OS (and config) level should result in a stable product overall. Just requires creating your own modules on top already available large software (and firmware) respository they have.

[@tristenr1293]

Please for the love of everything, build a good open source Linux firewall with a GUI. However, an alternative you might consider is building a web interface for VyOS if it suites your needs, or see if you can get in contact with the VyOS team to get yourself involved in their supposed GUI that is in the works.

[@tomazzaman]

Actually, this one seems to be favored the most in the comments so I'm literally downloading VyOS to try and build it for our hardware right this moment! 💪

[@tristenr1293]

​@@tomazzaman Also worth a note, they announced in June that they would be adding a "Stream" branch similar to CentOS stream. Not raw bleeding edge but not the paid LTS. Haven't seen anything since but a quick Google search will bring that up. I've been waiting for that one to really dig deep into VyOS.

[@Fortless]

Hi Tomaž, I would recommend not building your own routing OS for now. Instead, i would recommend making an existent routing OS that was CLI-only also have a GUI: VyOS. VyOS is a Linux router that is actively maintained by a large set of people. The only downside of it: no GUI, everything is CLI. They are heavily working on implementing the shiny new VPP dataplane, which can be offloaded to supported NICs. I would aim the software development of this whole project into funds and focus for both a) finishing up their VPP implementation and making VyOS work on ARM and b) making your own management interface or, maybe, even a full fork of VyOS with a Web UI. This method ensures that you have a stable baseline that did most of your work already by packaging and organizing all the different routing and management daemons under the hood. All you have to do is use VyOS's Python interface to communicate with its configuration and operational daemons. Good luck! Looking forward for more updates.

[@michaelfriderichsen5781]

Hi Tomaž the first time I heard about your project back when you posted the first videos about it. The first thought that crossed my mind was: This is the perfect board to run Ubuntu Core and build my own Router/Firewall from scratch.

[@shanemshort]

if you're considering linux, it might be worth seeing if you can port VyOS over to the platform. If you can do that 90% of the hard routing stuff is done and it'll give you an API to configure the networking.

[@M_Gargantua]

The big comment on accessibility I will add on the DIY Router OS - As its something I strive for in my own work. Every GUI click should map 1:1 with *some* sort of Command Line input. With the GUI you skip the parser on the input, but if you were watching the SSH/CLI at the same time it should at least echo the commands it just ran when you interacted with the GUI. Part two of good modern UX is to make a GUI keyboard navigable, either with hotkeys or good use of directionality so you can jump from one element to the next without taking your hands off the keyboard. Remappable hotkeys and separating the localization file from the hardcoded display also helps immensely. It may be a game but Factorio is always a good public facing example with how they do the UI under the hood and they publish details.

[@smilespray]

Tomaž, love your shouting. The software for this device absolutely has to be open and flexible with easy access to installing your own stuff using standard packages. No special compilation. Anything else is a dealbreaker. Go with an established OS. Anything else will limit the lifespan of the device. I have great trust in your hardware judgement, but I need to know that software support will be maintained by the community if (god forbid) your business fails or you decide to move on to your next shouting project. Also, please allow physical space for NVMe drives with cooling attachments. Heat kills.

[@wkm001]

This router is DOA, he just doesn't know it yet.

[@dozerd42]

Motor firmware person here: great simple explanation of PWM! For a practical application of how this works in DC motors, folks should try turning a switch to a lightbulb on and off very fast. The more "on" time, the brighter the light. The less "on" time, the dimmer the light

[@BirknerAlex]

Hey, have you looked into just using DPDK for packet processing with a normal Linux distro? That would be the most performant solution and also allows implementing things like DPI. Currently I am not aware that any consumer router is using it, its more used for things like DDoS protections which needs line rate speeds.

[@tomazzaman]

Glanced at it would be a better term. In fact, NXP provides support for DPDK, and I do plan to look at it more in depth in the upcoming months.

[@BirknerAlex]

By the way, if building from scratch, please ensure that every setting and function is API first. The GUI should be just a stupid API client - its always annoying if there is no easy way to automate things or not all functions are exposed in the API.

[@tomazzaman]

@@BirknerAlex Absolutely!

[@TheRich464]

Amazing update. Excited for every update. I love seeing how the sausage is made

[@tomazzaman]

Thank you! 🙌

[@SEDARTA_RECORDS]

8:28 - The only thing that I have found that works when it comes to fan cooling is to manually turn on and off a simple 2 pin fan running on 8.4V. It's the only thing right now that is 100% reliable.

[@b.wolthuis8178]

You asked not to directly critize the choice to not being able to load Opnsense on the device, however, for me up to this moment it was the main reason to follow this development…. perhaps I have to reconsider my point of view 😊

[@rpsmith]

I have been using and deploying pfSense for over 16 years and it can take a decade or more for a firewall OS to reach a stable, secure, and feature rich distribution. Also, it took me years to become some what competent supporting it so you can count me out on this project!

[@sebastian05000]

You can load it but it would be limited because of freebsd support in terms of the chip

[@rpsmith]

@@sebastian05000 -- yes, he mentions that in the video. I'll stick with pfSense on Protectli hardware.

[@tomazzaman]

Fair enough. I love OPNsense as well, but, well, no official support for our hardware :/

[@bigpod]

ARM has been here for years and if freeBSD cant make a good stake in that world sorry it just isnt worth our while maybe opensesnse needs to recosider its base like truenas did

[@rodrigofilho1996]

Pay a dev to add your hardware to OpenSense, then contribute the code. Your router will get more visibility and trust from the comunity.

[@Dr.Hoppity]

I’m in two minds about your router OS. You’re totally right we need a more “high end” Linux router distro cf openwrt, but it’s such a huge undertaking. I wouldn’t be confident using it until it’s proven, like openwrt and pfsense are

[@tomazzaman]

Yes, wholeheartedly agreed. This is no simple task.

[@FredrikRambris]

Vayatta or VyOS is an opensource os built just for this. If i remember correctly it has a commercial frontend.

[@glimps84]

If you're starting from scratch on the OS, please, PLEASE, OpenDaylight!!! Make it an SDN BEAST! A modern Linux based router with officiel modern sdn support would be life changing in my neck of the woods. Thanks for all the work

[@djenson]

You seem to be have to make many difficult choices purely because of the SoC you chose. Will be interesting to see how this develops

[@tomazzaman]

Comes with the job - and I love it. And also all of you who comment to help me steer these choices in the right direction!

[@autohmae]

I think I agree with some of the people in the comments, building your own will long term mean more work. My suggestion is to look at how to get someone to create drivers in for the FreeBSD kernel.

[@sanjikaneki6226]

HW: - man that is overkill with I2C busses you can probably get a rid of most those muxes if you properly manage addresses / sensor types +not so sure if all are usefull in actual production runs - depending on your OS you may even leave the eMMC MUX on the wifi ( if you can fit it all in RAM) - you CAN use 3 wire fans if you control them in voltage , tho this would require a free* smps or dedicated driver SW: -interesting trade offs BUT i think you can fir them all in the same eMMC and maybe leave some jumper somewhere for selection - regarding custom LINUX : it was about time someone did this ! the core (HW agnostic part) can be reused later and then accelerators can come on top is there tooling /sw that is open source for compiling linux for that line of NXP MPUs?

[@Gersberms]

Sunon makes quality fans. Good choice. It doesn't all need to be Noctua!

[@GabrielM01]

Another player in the router OS space thats based on Linux and isnt so focused on being light on resources? im all for it man, props to you

[@jakastermenszky7209]

OMG, that’s a really bold move. Thumbs up!

[@TrTai]

Man kind of a bummer on the OPNSense front, hopefully FreeBSD gets some more love to make things more ARM friendly in the future. Rolling your own is going to be rough, but could be interesting. VyOS like others have mentioned seems like the next best bet otherwise being Linux based and (in theory) should work with the hardware, otherwise just building ontop of Linux or OpenWRT would probably be the way to go imo. I think Open Sourcing down the road is fair, keeps from undercutting your business while it gets off the ground, while still giving back to the community and enabling them to do more cool things. Looking forward to seeing more on the Founders edition, looks sweet! and Looking forward to seeing what comes next!

[@JasonFritcher]

Have you looked at the difficulty of porting or writing a driver for the network accelerator on the NXP chip? How would the effort to do that compare to assembling a new distro from scratch?

[@Ixsys]

I was so super excited with this project when it was using off the shelf matx boards and was being CNC'd out of Aluminium. I'd have pony'd up for just the case in very close to a heartbeat. Probably would've bee interested in a pre-made system using the same case. Now, I gave this a good shake, and have been following for months. I have serious misgivings about this project now. Building your own OS is a fools errand. Do not go down this route. Focus on delivering the product that can be run with a wide variety of OS's. When and if you succeed with the first project you can do a 2nd gen, and examine the OS aspect of a future project after you have some iterations under your belt. Do NOT start using proprietary parts like 35mm fans. This is not the way. Stay with standard sized fans, so as not to stifle those who are going to mod this, and do fun projects to build on your own starting point. It also means that if something breaks or fails, a user can DIY a fix instead of being forced to buy unicorn parts that come from only one source. Seriously I cannot stress this enough - DONT USE PARTS THAT ARE ANYTHING OTHER THAN EASY TO GET. This custom board... it sounds really cool. Honestly it does. But warranties, patching, updates etc are all going to be on you now. This is a massive amount of work and again is a bad idea. Custom Case - Super cool idea. Now make a version for people who want to use off the shelf matx/ mini itx boards, with a pcie slot, a 1U form factor PSU, and some sort of mounting hardware so you can utilize direct cpu contact and heat disipation though the casing. I think Zalman did a fanless case, and Supermicro offers something similar in the Embedded solutions section. I fear that where once I was full speed ahead on this project, I now forsee a flash in the pan product that will not take hold or have much longevity. I hope I'm wrong. I'll wait till next update to see the case and such, and then depending on what I see - probably go a different direction for my firewalling. Best of luck to you and your team.

[@alec1575]

I love these monthly report vids, also, 1,500 total parts is more than I expected

[@snarkywombat155]

My opinion for what it is worth. Why reinvent the wheel? It seems that you're going about it all backwards. If you are wedded to ARM, it feels like going it from scratch seems wasteful, let alone a massive rabbit hole. If it were me I would absolutely use a BSD core, hell you could even fork OpenSense or some other OS. Even if none of that is a concern, my Cyber Sec brain immediately turns to CVE patching response, zero exploits and all manner of other leverages that would crop up that you'd have to fix, assuming you even knew about them.

[@rabiddawgr]

And then you spend all your time developing kernel code to handle all the offloading hardware. As opposed to Linux where all that code already exists. I mean ARM architectures aren't even officially supported, there isn't a massive community dedicated to maintaining an arm core. This would hit you right into the CVE patching response, because you don't have a OEM or wider community targeting the architecture.

[@EtienneBAuger]

A great approach for deploying and preconfiguring the router is to allow configuration through declarative config files that sync with the UI, and vice versa. I really appreciate how NixOS handles this concept! 😃

[@thamescomputersanddatarecovery]

Happy to hear your Co-Founder had his holiday in Kenya.

[@stevefxp]

Have you talked to Noctua about their fans? I use their fans in my cigar humidor, so I can have more granular control.

[@tomazzaman]

I have not, not reason other that I consider them more a consumer brand, which means they likely can't match Sunon's prices (for the same level of quality).

[@zickzack3106]

Hey! I have only stumbled upon this video by chance and don‘t know all the context, but wanted to strongly recommend „Noctua“ from Austria/Taiwan for the actual best fans point stop for your high-end router. They have a 40mm fan and I think they do custom cooler design too. They are a community favorite for performance, longevity + support and the T30 you showed is kind of based off from Noctua‘s Chromax A12x25. Take care

[@oa_munich]

JunOS is based on FreeBSD, and not the most current version too. They decided it was easier to lock down the OS and the maintain drivers

[@BrianG61UK]

I hope this isn't going to be another router that is advertised as gigabit capable, which can only get anywhere near this if you do everything with the data path acceleration architecture. But if add additional real world requirements like QoS, it becomes a router that struggles with speeds much over 20Mib/s.

[@KD_Puvvadi]

I don’t think mass is bigger factor for heat dissipation. It’s surface area that matters in thermal engineering

[@bastian775]

I think it's great progress. If we require fans "13-in-a-dozen" 4-pin 40 mm fans would be nice. I somehow remember calling sunon "rattleon". Software is indeed a problem since the hardware offloading does not work. If the ISP like my ISP uses horrible legacy PPPoE as quite a lot still do you don't get anywhere near full speed this really is a problem with PPPoE >2.5-3 gbit. Also for many people OpenWRT will do, maybe we can still have it as an option, especially if we can just use stock OpenWRT without too many adjustments so we keep receiving updates or can compile a new one without too many problems/minimal adjustments.

[@ss-xy2im]

Currently no decent Linux router/fw solutions exist, my router is Ubuntu with iptables, isc-dhcp, fail2ban & crowdsec, all controlled from the terminal. It's high time for a decent all in one linux router with webgui.

[@ernstoud]

But does NXP (and the other chip suppliers) release Linux source code for all the SOC / IC functions? So that you can build them yourselves? I am a bit disappointed it is not going to be OpenWRT since my router runs a number of packages available pre-compiled for OpenWRT such as the RIPE Atlas probe, DigiTemp a.o.

[@tomazzaman]

Oh no, don't get me wrong, you will be able to install Linux on the device and run pretty much whatever you want! So regardless of what we decide to ship ip with, you'll have completely free hands with it!

[@TheHeidarsig]

Vyos or even frr as source and you can spend time to do the front ui... then you have all the routing fuction needed

[@tomazzaman]

This is exactly what we're testing these days! 💪

[@funkiam9214]

frr +1 VyOS is not fully opensource only night build

[@rfitzgerald2004]

It's refreshing to hear your experience with ADHD and not wanting to take time away from your passion projects, I'm exactly the same! I've only had the one period of burnout so far though! As for number of components, I'm guessing around 500 including things like the power module and all the SMD chips etc

[@rfitzgerald2004]

Wow 1500! 🤯

[@kylechase4621]

Yes to VyOS, been using VyOS and Vyatta forever, and I love it. You can also build tailscale into it with vyos-modular

[@madnessandescapism]

I love your approach with building firmware yourself and I think there is a market for another opensource router OS. I'm sure you will be able to pull it off. As for schematics - this already is a big deal. I do not think that publishing PCB layout even makes sense. If I want to tinker with it - schematics would be more than enough, schematics + step files = even more awesome!

[@danbrown586]

As to the OS--what about VyOS? Linux-based, so it should be able to take advantage of your CPU's features. The routing software is already done. What it lacks is a GUI. So rather than building a Linux-based router OS from scratch, your development time could be spent on a good GUI front-end for what's already reputed to be a solid routing platform. Admittedly, I say this with no direct experience with that project--I use OPNsense--but it seems it'd give you a huge head start compared to starting from scratch.

[@hungnguyen-pl3eg]

He's trying his best to create a circuit board that any manufacturer in China can easily produce, one that will undoubtedly destroy his board in terms of performance, stability, and efficiency, at a price that's only 1/10th.

[@Luisa_adams]

Watching from kenya... That was giraffe centre where you guys visited

[@collectorguy3919]

A less painful approach is to create a spin/fork from an existing Linux distro that has an active community doing the heavy lifting. For example, Alpine Linux would make a lightweight modern Linux base, and the web UI would be a clean slate. You'd still have a lot of work ahead to integrate NXP's kernel with your secret acceleration sauce. Your 'community edition' would not be accelerated. The fastest way would be to focus on the hardware, and support OpenWRT, but 'fastest' doesn't appear to be your objective.

[@kanguruster]

Fantastic video! Yay, holidays! Yay, caring for family! Yay, avoiding ADHD burnout! Then there's the technical content 👌

[@dsfdfssqdfsqfdfqsd4235]

Real nice, tyvm !

[@frankle_au]

If your gaming performance and buffer bloat can beat the Ubiquiti ER4 (when setup correctly) then I’m onboard. Looking forward to see where this ends up!

[@Daniel15au]

Why don't you reuse the OpenWrt backend and build your own frontend? The backend and frontend aren't too tightly coupled.

[@AhmadSwailem]

Well this is interesting 🤔 Hope you crack things out.. Best of luck 👍

[@antonk1620]

why not alpine linux? its small (can fit in several mb and run from memory), secure and well supported, and not castrated like openwrt. its even originally was purposed for routers, but looks like never used on them (i think i even seen some gui packages for this purpose)

[@GabrielM01]

In terms of the open source of the hardware, why dont you goes make your own license that protects you from people copying commercially the router while still giving us the schematics? much like what Futo does for the GrayJay app, making it more Source Available but thats enough

[@ZoeyR86]

You can take the mass of that and measure the temp rise to calculate the thermal input

[@elib3544]

What is the custom OS based on?

[@vgx-sam8164]

Keep going, following you from your router hacking videos 💓

[@pauliusruginis5264]

As someone who work for the biggest IoT mobile router manufacturer on the market. Biggest advice is, take OpenWRT or OPNsense or PFsense and run it. We have a huge team 100+ people working and it still takes a year to launch a product to a decent working state by running an existing OS solution. Building and OS will take tens of thousands of hours even for the very basic stuff that cheap Chinese routers are able to do. Not talking about any other features like VPN support that will be able to take advantage of the hardware that you have put in. If you don't like OpenWRT WebUI, scratch it, make something of you own Even building the UI for exiting OpenWRT stuff will be a pain in the but, talking from experience

[@GlenShok]

Try Netgate TNSR. Seems like you are re-inventing the wheel here.

[@EraYaN]

Making your own OS is probably a bad idea, better to start with something like VyOS or maybe even TNSR. Or just customize OpenWRT a bit, there are already 3rd party UIs for it.

[@urban6989]

why not yocto? it's much more flexible than buildroot, happy to talk more about why it's better option than buildroot 😉

[@srus]

Se strinjam. I agree.

[@tomazzaman]

Simply because I have more experience with Buildroot and can achieve similar results.

[@matejsokolec2973]

@@tomazzaman I wanted to ask this as well. I was a Yocto dev for 5 years in Croatia, so next door, and in case you need some help maybe jump in to help.

[@OrcunKurnaz]

If you are planning to create ur own ui, why not using vyos? As I know they can run on arm.

[@xdevs23]

16:20 maybe NixOS or postmarketOS could be something to consider.

[@nunoduarte9805]

just because we expect to have a custom linux routing software I would pay 10% more for the device

[@streamx3]

Please do not opensource the PCB if that was what you ment. Boardview is an option, if you want to help repair technicians. But anything beyound that will not be used in a good fate. At least thats my experience.

[@Sithhy]

Would be really cool if you were to go with Alpine Linux for your router as it was made for this purpose & I have yet to see a router that uses it as its OS. As for the fans, please go for 40mm as they're much more standard size than 30 or 35mm so getting a replacement in case it would fail would be much easier

[@victorshane4134]

Mikrotik has Installation Media available. I hope your router will be able to run it! :D

[@AmauryJacquot]

as a suggestion, you could opensource the PCB once you have made a decent amount of money out of the projet (say 5 to 10 times the development costs)

[@jordantekelenburg]

Are you going to release the schematics?

[@ComeCaramelos_]

- (unfortunately) "[...] companies don't really invest that many resources into developing firmware for it" - Proceeds to not developing firmware for it 🙃

[@lohikarhu734]

Hmmmm...as an old embedded systems guy, your controller sounds rather complex... 1500 components is more like a laptop... I'd love to see your schematic and your aims for the product?? Not assuming that you are not competent, but, having done this stuff for 35+ years, and worked in both consulting and very high volume environments.

[@jirkasvitil2762]

You probably thought about that. But isnt it possible to use existing distro and just add packages on top of it? Seems easier than mantaining full distro. You could just maintain the package sources

[@bigpod]

the fact that they are going from build root onward kinda shows how bad they have it they should probably base on something like fedoras silverblue or properly ublue base as it would get them all features but they wouldnt have to maintian WHOLE of RootFS or packages plus it would come with immutability and such

[@berlinberlin4246]

Stick to openWRT and get an custom UI merged with the main openWRT

[@alcorza3567]

What thermal interface are you going to be using between the component and the case itself (if anything). A regular thermal pad, paste of some sort? Or have you considered something like the Honeywell PTM7950 pads instead?

[@nazar5549]

hey, is all that hardware acceleration stuff supported with an open source kernel or it's a proprietary blob? does NXP intend to merge the patches into mainline? I'm fine with any distro, as long as community will be able to archive full performance without signing NDAs or something similar.

[@StephenPreston]

I would be interested in the using the software on my own hardware and would be willing to pay a subscription once its finished to support dev ..... I really like my pfsense but I have been looking for something to swap it out mostly because of the limitation of freebsd

[@j2klegend]

Super excited! Your energy gives confidence for all 👍

[@rpsmith]

"Super excited! Your energy gives confidence for all" -- Even if his lack of critical thinking doesn't!

[@j2klegend]

@@rpsmith As per my understanding he always cover all possible risk factors. Sometimes Nike's slogan do the work.

[@tomazzaman]

No risk, no fun! :)

[@Tofflus]

This project broke my heart when ARM CPU was chosen. I had hope it was gonna use x86. I won't be able to use the software I use. :(

[@rpsmith]

You are 100% correct!

[@tomazzaman]

Fair enough. I don't like it either honestly. (The vendors not providing drivers for FreeBSD that is)

[@bastlub]

​@tomazzaman, i think your response is beside the point. The problem here is that this project is based on an ARM cpu with a custom extension, which is not supported on ANY os. There is only drivers that you need to glue-patch into an os

[@yoshidis4]

The best fans are Noctua. Nobody considers otherwise. Period.

[@bigpod]

Only reason people talk that way about noctua that way id because their pc fans are good and somewhat silent but in induatrial space they are juat one of the players and its hard to say which is actually the best

[@c0p0n]

Mate, what openwrt do with the UI is modern frontend development. That's how most websites work since 2015.

[@ksenchy]

Why would wifi need to communicate with emmc? Surely you aren't writing anything to it? The problem is that emmc is a flash storage and it has a limited write cycles

[@jordanwardle11]

Would a fork of vyos with a custom gui (plus other required changes) work? Or would you rather a fully custom os?

[@Pistolshot]

Amazing work, explanation and video. Learning a lot. I have question: Could this run a proxmox or vmware exsi?

[@floodo1]

I doubt many ppl consider a custom OS much of a benefit and most would consider it a detriment (not the same as what they are already familiar with and known to work well) Id do the work to port/fix a known good popular OS rather than roll own