Your Router Sucks. Build Your Own Instead! (SPONSORED)

[GodAtum]

This is the most stupid thing I've ever watched. Most ISPs combine the modem and router so it cant be replaced. A lot of people in this thread don't seem to understand either.

 

How is LTT expecting me to connect my self-built router to my telephone line once I throw away the crap ISP modem/router? the video NEEDS to specify I need to buy a separate modem that my ISP supports.

[Crass Spektakel]

https://i.imgflip.com/703u8r.jpg

 

I am from a different time when you had to build your router ALWAYS yourself or not have any router at all. My first router in 1994 was a 486@80Mhz with 2x100MBit Ethernet and six (!!!) RS232 connections running CSLIP at 115200bps because Ethernet back then was expensive and until 2015 all my routers were build by myself, running a heavily modified Debian and later Ubuntu. My last DIY router was a Core2 with 2x3600Mhz, 8GByte memory, Geforce 8800, 5x1000MBit, 5x3000GByte HD (RAID5). And here is the trick, it did the routing stuff as a side job. Its main job was being a decent server. Therefore it didn't matter how much power it draw. And honestly the server was MASSIVELY overpowered even though it ran SMB, SQL, Apache, Squid, Teamspeak and myriads other services. Oh, did I mention that I often logged locally into the system, working on two screens, coding stuff or playing Left4Dead? And still the VPN was working at full speed of the internet connection of 100MBit back then.

 

Honestly for a mere router I would rather suggest using a Pi, not a PC. Most likely even a Pi Zero would do the trick.

 

The presentation itself was fine. Not too much to get bored but enough to get the idea. And tbh I didn't even knew that OPNsense made it so far.

 

But the overall idea of why to use an DIY router and the implications were done very poorly. Face it, a brand AIO SOHO router costs €50-€200 and uses 3-10 Watt. Separate the pieces (modem, switch, router, splitter etcpp) and you are closer to 30W - WITHOUT THE PC. To make matter worse, raw modems are usually as expensive as routers and need the same power and in many cases they ARE actually routers switched into modem mode. No joke.

 

Using an old PC as a router is a VERY bad idea.

 

But adding the feature to an existing Server, NAS or other appliance? Not a bad idea. In my case it was a simple script which made the PC into a full featured Router with excellent firewall - see attached file. That is actually the whole magic of a router and a firewall. And a pretty fast one too.

 

packetfilter-example-20150514

[Sorachan]

8 hours ago, mynameisjuan said:

Can you clarify on what packet sizes you're testing? Because 1-2Mpps is far beyond consumer grade firewalls and even routers at 1500B.

I mostly tested ASUS, Xiaomi et al higher end routers, they are very close to commercial grade products, that's why I think they have exceptional value proposition (a lot cheaper but still great performance). Of course many lower end consumer routers are nowhere near this rate, but still definitely better than most DIY routers.

 

99% of consumers do not use their routers as VPN or ad-blocker, CPU bound routers have no advantage over these well designed consumer routers that I can possibly think of, unless you run VPN or use them as a NAS (but why should you? unless you live in China or Iran). That's why whenever I see video from LTT stating these are better than normal routers I have absolutely no idea what they are talking about...

[iLikeBananas]

18 hours ago, Senzelian said:

We might have some of the worst electricity cost around the world here in Germany, but at least the German company AVM makes pretty damn good routers.

I disagree. The AVM 7590 is a router from 2017 and it still costs 200€. I had the 7490 before I upgraded to this one, and the 7490 would constantly crash. At least for me. I would love to replace this with a DIY open source router.

[EasyGoing1]

On 11/8/2022 at 10:13 PM, cowsgomooo said:

This video was pretty sloppy and poorly articulated

 

Doesn’t even explain why you’d build your own router over buying a nice aftermarket one.

Yes, he does at the beginning of the video. This isn’t something your average user would ever do.

 

On 11/9/2022 at 3:24 PM, GodAtum said:

This is the most stupid thing I've ever watched. Most ISPs combine the modem and router so it cant be replaced. A lot of people in this thread don't seem to understand either.

 

How is LTT expecting me to connect my self-built router to my telephone line once I throw away the crap ISP modem/router? the video NEEDS to specify I need to buy a separate modem that my ISP supports.

When you purchase internet service, all ISPs will accommodate you if you want to use your own router. Just say “I’m using my own router” when you buy the service and they’ll set you up accordingly with a nice Ethernet Jack ready to go … DOCSIS modem, Fiber bridge, DSL modem (depending on where you live), last mile point to point WiFi, Satellite … lots of possibilities out there depending on your isp. Combining the modem and the router is a relatively new practice among ISPs and the majority of people on the planet don’t have access to the latest and greatest service either.

 

On 11/9/2022 at 3:24 PM, GodAtum said:

This is the most stupid thing I've ever watched. Most ISPs combine the modem and router so it cant be replaced. A lot of people in this thread don't seem to understand either.

 

How is LTT expecting me to connect my self-built router to my telephone line once I throw away the crap ISP modem/router? the video NEEDS to specify I need to buy a separate modem that my ISP supports.

No the video doesn’t need to specify, you need to educate yourself, or better yet, don’t build your own router.

 

On 11/8/2022 at 10:31 PM, Senzelian said:

We might have some of the worst electricity cost around the world here in Germany, but at least the German company AVM makes pretty damn good routers.

But your bandwidth in Germany is crap too from what I understand … like 15 years behind the rest of the planet… what’s up with your country? Why do you guys hate Internet service so much? 

 

On 11/8/2022 at 11:18 PM, DrMacintosh said:

Sure, you could built a giant home router that consumes like 100W+ and requires monitoring and tinkering, or you could get 2 really nice Mesh routers from a brand like Linksys or Ubiquity that consume less than 20W combined and cover the majority of home networks and take 10min to setup. 

Why would you use two routers? Or are you talking about WiFi access points? If you only have one internet connection then you can’t use two routers.

 

On 11/9/2022 at 12:49 AM, Arika S said:

this seems massively inefficient. not in only power draw, but also the amount of powerpoints you now need.

 

not everyone has a server rack in their basement, so i don't know where you really expect "normal" people to put these where it isn't going to look massively ugly.

Your comment would make sense if this video suggested that this option was the best option for everyone … this video is for a very small group of people who understand what they can do with a setup like this … this offers maximum flexibility for a network and can provide features you’ll never find in a Netgear … such as multi homed VPN end points, limitless DMZs, bandwidth shaping to exponential levels, traffic logging to any level of detail your heart could desire, intrusion detection and alerting, block entire groups of public IP addresses (Nigeria? China? Russia?) possibly even IP geo fencing for faster gaming experiences … point is, the possibilities are endless as opposed to a typical home router where you can do … only what they let you do and that’s it.

 

On 11/9/2022 at 3:36 AM, Sorachan said:

LTT should do better.

 

The performance and stability of a router cannot be measured easily without professional tool. There are industry standards like RFC2544 for a reason.

 

You should compare dedicated routers and CPU driven routers by aggregated small packet results. Unless you use modern power hungry CPU, dedicated routers are really hard to beat.

 

If LTT truely wants to be professional reviewer, you guys should consider investing a IXIA optixia XM2 or similar IP performance tester. Dedicated routers exist for good reasons other than power efficiency. From my past testing, most DIY routers can do about 100-300 Kpps, while most consumer grade routers like can do 1-2 Mpps, newer routers like AX86U perform even better.

But commercial routers are nothing less than typical PC hardware designed only for routing with some kind of Linux OS on them. Unless we’re talking Cisco or even Microtik which makes a pretty darn good router at a low cost … though I would love to see a comparison of off the shelf home routers compared to a solution like this under heavy load testing … stream up a few COD sessions and some VOIP streams and some video conferencing streams and see how all that traffic is handled in real time. Heck the majority of cable internet is still asynchronous for that matter … the average home user has no concept of what a solid high speed synchronous internet connection feels like… 

 

On 11/9/2022 at 8:12 AM, mynameisjuan said:

 

RFC2544 is great for testing forwarding rates for IMIX, but it also doesn't paint the whole picture. That is why Y.1564 and supplemented by RFC 6349 have proven give closer to "real world" results.

 

 

To break it down further and to clarify, it also needs to be separated between routing and firewall as there is a large difference in performance between the two. I'd say include switching but at the consumer level, modern switchchips/ASIC are cheap and 99% of the time can do line-rate at 64B. When it comes to DIY, there is a lot in variability to make the results meaningful.

 

Pure routing in hardware vs software is not even a competition unless you begin to throw CPU at it. But as NOS' improve kernel level support along with software and NIC's supporting more HW offloading, the difference in routing performance between dedicated consumer routers and DIY routers is shrinking. However, power/performance still doesn't make it worth it.

 

Now when talking about consumer/DIY routers, they are a combination of routing, switching and firewall. It's not until the mid to upper end of enterprise firewalls where even basic L4 connection tracking can be done in hardware. Consumer/prosumer firewalls will almost always have the CPU involved for conntrack. This results in bringing the performance between consumer "routers" and DIY much closer depending on CPU because you're losing the benefit of HW routing and punting to a much weaker CPU. That said, consumer routers can have the ability to HW offload EST connections, reducing the load on the CPU.

 

While I would like to see the labs run various performance test, in the end I think it's mostly a waste of time. Too many variables, too little of a target audience and I don't have faith the test would be executed or interpreted properly. If the do I would like them to focus on the DIY side of thing and power consumption that way they can demonstrate how wasteful they are for 99% of people. Unless a user is looking at some NGFW features, it makes more sense to just stick with consumer routers. We're talking 10-20w stressed vs a DIY that idles at 40w, let alone has any significant load on it.

 

Can you clarify on what packet sizes you're testing? Because 1-2Mpps is far beyond consumer grade firewalls and even routers at 1500B.

Sometimes it’s not always just about the performance …. I can’t tell you how many times I’ve wanted better traffic logging in my Netgear router only to be oh so disappointed with what their logs capture not to mention they don’t store much history. Also, try and set up a nailed VPN connection to a vpn provider of your choice and your out of luck with a Netgear you have to use their brand of vpn service or nothing at all … then what if I want my virtual machines packets to be restricted to only use the tunnel while other devices don’t … flexibility is where it’s at with a diy and there’s no way to compare against home routers in that space … at all!

[DrMacintosh]

1 hour ago, EasyGoing1 said:

Why would you use two routers? Or are you talking about WiFi access points? If you only have one internet connection then you can’t use two routers.

Ever heard of mesh Wi-Fi? 

[Arika]

5 hours ago, EasyGoing1 said:

Your comment would make sense if this video suggested that this option was the best option for everyone … this video is for a very small group of people who understand what they can do with a setup like this … this offers maximum flexibility for a network and can provide features you’ll never find in a Netgear … such as multi homed VPN end points, limitless DMZs, bandwidth shaping to exponential levels, traffic logging to any level of detail your heart could desire, intrusion detection and alerting, block entire groups of public IP addresses (Nigeria? China? Russia?) possibly even IP geo fencing for faster gaming experiences … point is, the possibilities are endless as opposed to a typical home router where you can do … only what they let you do and that’s it.

Your comment would make sense if the video and title suggested any of what you said.

[mynameisjuan]

10 hours ago, EasyGoing1 said:

But commercial routers are nothing less than typical PC hardware designed only for routing with some kind of Linux OS on them

This is not true, even at the consumer level. Sure, switchchips/ASICs and their feature set have trickled down and made their way into PC hardware, they are most definitely not "PC hardware designed only for routing". Server/PCs having the ability to route/switch at an acceptable level is still a relatively "recent" offering and that is only because this trickle down of hardware.

 

From consumer to enterprise hardware, the same designs are carried throughout. You have a data-plane which consist of the switchchips/ASIC's to handle the actual forwarding in hardware and you have the control-plane which handles the computation aspect and would be the "PC hardware" aspect to your comment. This design exist across the board because switching/routing/stateful FW (from hardest to easiest) is difficult to do in software. This decoupling is what allowed us to achieve a level or performance that accelerated much faster than CPU's could ever handle and keeping it in a power efficient package.

 

The largest factor in why DIY firewalls are able to compete with consumer/prosumer hardware is because state which as I mentioned previously involves the CPU and levels the playing field. If you wanted to more than basic forwarding with DIY, you need NICs that can do so in hardware. Essentially, its not PC hardware redesigned for routing, it's about configuring a PC to be more like a router.

 

10 hours ago, EasyGoing1 said:

Sometimes it’s not always just about the performance …. I can’t tell you how many times I’ve wanted better traffic logging in my Netgear router only to be oh so disappointed with what their logs capture not to mention they don’t store much history. Also, try and set up a nailed VPN connection to a vpn provider of your choice and your out of luck with a Netgear you have to use their brand of vpn service or nothing at all … then what if I want my virtual machines packets to be restricted to only use the tunnel while other devices don’t … flexibility is where it’s at with a diy and there’s no way to compare against home routers in that space … at all!

Well yeah, once you are able to involve software, of course there are no restrictions of what can be done. But that's not my argument.

 

My argument is that the power efficiencies and hardware level performance are traded for software features and it's too commonly recommended to go the DIY path when many that do make no use of most the features. Yes, if you know what you want and can only be had at a reasonable cost with DIY, go for it. But LTT's video never demonstrated any of it.

 

Take a Mikrotik hAP ac2 for example. $70, great performance, deep feature sets, sips power and in a package smaller than most phones. Not only that, unless you're looking for IDS/IPS, it has the ability to configure everything you mentioned thus far and can be more granular. Not only that, their latest revision introduced containers that can be ran natively...but with AC2, memory can limit that.

 

DIY doesn't make sense for many and not worth the trade offs.

 

11 hours ago, EasyGoing1 said:

the average home user has no concept of what a solid high speed synchronous internet connection feels like… 

Counter point; The average user that has access to and claims a massive improvement in real world usage also over estimate their claims and/or associate it with the increase in throughput.

 

If the same medium was used, the average home user would not be able to discern between a 100/10mbps and a 1/1gbps connection. But if you have them compare a 100/10mbps COAX vs a 100/10mbps fiber connection, the average user would notice a real world difference in everyday use due to difference in latency.

[luckybob77]

Quote

Server/PCs having the ability to route/switch at an acceptable level is still a relatively "recent" offering and that is only because this trickle down of hardware.

YEA.....  NO.  

 

I've ran my own router since "high speed" internet was 1.5mbit.  It was a simple task to use a 486 in that era.  I've had literally zero issues with home made routers.  Granted anecdotal evidence being what it is...  My favorite and longest lasting was a Smoothwall based unit that ran on a pair of Intel pentium 3 xeon processors.  Had no trouble with speeds up to about 500mb/s.  above that, the old parallel PCI bus just couldn't handle it.  My latest unit is a bit overkill:  https://www.ebay.com/itm/143586569305  That's a router with some chest hair!

 

I still have that 486 and the p3-xeon.  

 

Quote

compare a 100/10mbps COAX vs a 100/10mbps fiber connection, the average user would notice a real world difference in everyday use due to difference in latency.

again, no.

 

your reasoning is flawed.  ALL things being equal, there is no difference between fiber and coax.  but things arent equal.  the differences someone would "notice" would be much more dependent on the company running those services.  Most coax is only coax for the last mile or so.  Most everything else has long since been upgraded to fiber.

[mynameisjuan]

14 hours ago, luckybob77 said:

YEA.....  NO.  

 

I've ran my own router since "high speed" internet was 1.5mbit.  It was a simple task to use a 486 in that era.  I've had literally zero issues with home made routers.  Granted anecdotal evidence being what it is... 

I never said it couldn't nor have been done until recent, I said at an acceptable level.

 

Depending on the time frame of 1.5mbps, routing was still being done in CPU, forget basic L4 FW and any chance of HW offloading for conntrack. I'm talking on enterprise/carrier hardware let alone consumer hardware. It was not possible early on to route in hardware. So yeah, if you had the CPU to do so, it could have been done easily on consumer hardware.

 

I mentioned that throughput rose at rates which exceeded what CPUs could handle and it impacted FWs the most because routing, NAT, state, etc, is all running on top of it. Once chipsets were able to offload conntrack and/or memory for a few dozen routes, this reduced the impact on the CPU and limited it to just the session establishment.

 

Your argument and referencing server grade hardware actually backs my point. It was able to achieve those rates because it had hardware that could offload and reduce the cycles needed. This hardware came from the chipsets developed originally for network vendors which became extremely cheap and trickled down to enterprise then consumer hardware. This combined with current CPU IPC is why 1gbps can easily be done on a single core in modern hardware.

 

Im not saying it could not have been done this whole time. My argument was against the notion that "routers are just PC hardware developed for routing". You could always just throw more CPU at the problem to overcome it but would still be limited by IPC.

 

14 hours ago, luckybob77 said:

again, no.

 

your reasoning is flawed.  ALL things being equal, there is no difference between fiber and coax.  but things arent equal.  the differences someone would "notice" would be much more dependent on the company running those services.  Most coax is only coax for the last mile or so.  Most everything else has long since been upgraded to fiber

 

I was comparing access mediums alone, not all the variables between SP. I'm fully aware of COAX being last mile only but you're ignoring the fact that COAX has built in noise correction that introduces latency intentionally which does matter more for the average user vs raw bandwidth.

[EasyGoing1]

On 11/10/2022 at 7:13 PM, DrMacintosh said:

Ever heard of mesh Wi-Fi? 

Yes, but WiFi meshing is NOT routing… at best it could be classified as bridging but even that would be lower in the OSI model.

 

On 11/11/2022 at 6:14 AM, mynameisjuan said:

Counter point; The average user that has access to and claims a massive improvement in real world usage also over estimate their claims and/or associate it with the increase in throughput.

 

If the same medium was used, the average home user would not be able to discern between a 100/10mbps and a 1/1gbps connection. But if you have them compare a 100/10mbps COAX vs a 100/10mbps fiber connection, the average user would notice a real world difference in everyday use due to difference in latency.

I agree with this the ‘latency’ aspect 100% as latency is at the core of discerning the difference. Though I disagree with your generalization where the average user wouldn’t discern the difference between 100 or gig … though that is generally true until they need to download the next Call of Duty update - sometimes as large as 30 gigs in size where they would definitely feel a difference between100 and gig. 

 

On 11/10/2022 at 11:31 PM, Arika S said:

Your comment would make sense if the video and title suggested any of what you said.

Well it would be silly to include specific use cases in video titles, would it not? And is it unreasonable to assume that viewers will be aware of a video’s relevance in their world where they can just move along to something else at their discretion?

 

On 11/11/2022 at 6:14 AM, mynameisjuan said:

DIY doesn't make sense for many and not worth the trade offs.

Perhaps one aspect you’re overlooking … people who like DIY have other personal rewards that they gain in accomplishing the task … personal satisfaction of actually doing things themselves … which usually is a partner to ‘knowledge gained’ kinda thing.

 

But I do need to look at Microtik … any idea if they offer geofencing for online gaming in their routers? Cause OPNsense has plugins for that.

 

On 11/11/2022 at 8:13 AM, luckybob77 said:

your reasoning is flawed.  ALL things being equal, there is no difference between fiber and coax.  but things arent equal.  the differences someone would "notice" would be much more dependent on the company running those services.  Most coax is only coax for the last mile or so.  Most everything else has long since been upgraded to fiber.

I will completely disagree with this … coax will always have the DOCSIS overhead in the link where it needs to modulate a carrier on the same copper that Carrie’s television signals and a host of other signals. Fiber has no such overhead. Also, wave propagation along copper with waves being electromagnetic in nature and also where waves on the same coax are in chronic simul-state existence and their properties constantly being modified by other waves on the same copper will also introduce complexity into the physics of the circuit which adds to the overhead. Again, Fiber has none of those variables to deal with. No electromagnetic waves to deal with at all, and a single strand of fiber can have true non-interfering simultaneous “conversations” happening since different carrier waves can be tuned out of spectrum by altering the frequency such that the photons never interfere with the other photons in the fiber. And a properly engineered fiber network can absolutely be perceived as feeling a lot more “snappy” when used in comparison to a coax network … I guarantee it.

 

On 11/11/2022 at 6:14 AM, mynameisjuan said:

This is not true, even at the consumer level. Sure, switchchips/ASICs and their feature set have trickled down and made their way into PC hardware, they are most definitely not "PC hardware designed only for routing". Server/PCs having the ability to route/switch at an acceptable level is still a relatively "recent" offering and that is only because this trickle down of hardware.

 

 

I was actually being more generic in my assertion - where a router will consist of a CPU, a BIOS of some kind facilitating boot up sequences, a static storage medium of some kind, RAM and standard input output.

 

You brought up some excellent points though and also managed to answer some questions I’ve had for a while concerning general router design methods and how we have worked towards improving packet latency in the router itself which traditionally was a bottleneck that was always just there and we had to accept it. Cisco had some creative buffering that they did in IOS that mitigated the serial nature of bit processing in CPUs in general … but what you described sounds like the whole design paradigm for routing hardware has completely changed in the very core of how it is even conceptualized … where today they seem to be designing from the perspective of the packet and it’s path through the router and then engineer the hardware to be as accommodating to the packet or packets as possible even adding off cpu processing that makes the whole system … “flow”

 

good stuff

[DrMacintosh]

1 hour ago, EasyGoing1 said:

Yes, but WiFi meshing is NOT routing… at best it could be classified as bridging but even that would be lower in the OSI model.

The master node does perform routing....

[Arika]

1 hour ago, EasyGoing1 said:

Well it would be silly to include specific use cases in video titles, would it not? And is it unreasonable to assume that viewers will be aware of a video’s relevance in their world where they can just move along to something else at their discretion?

On 11/11/2022 at 8:31 AM, Arika S said:

Your comment would make sense if the video and title suggested any of what you said.

 

 there was nothing in the video that said it was for more advanced users. in fact the entire video was showing how easy it is and to breakdown the misconception that it's only for advanced users buy going will the cheapest parts and easiest software to use.

 

so unless you can timestamp in the video where thy said;

On 11/11/2022 at 2:46 AM, EasyGoing1 said:

this video is for a very small group of people who understand what they can do with a setup like this

 

then my points still stand

[LogicalDrm]

3 hours ago, EasyGoing1 said:

good stuff

Hello. Several of your posts have been merged (14 total to 2). We like to keep forums neat and with less clutter where it can be avoided.

 

There's feature here that allows you to easily reply to multiple posts in one, multiquote. So you can select all posts you wish to reply to by selecting them with the double arrow button on the left of the row. Then click the box appearing on the right bottom corner to write your replies. On desktop, you can remove extra quotes by selecting element (top left corner of the element) and delete key. You can also break long posts to several quotes with double-enter (double-line break).

 

Thanks.

[Pepino-CH]

Its great - BUT. I use at home NAS-server Qnap TS-677 do you think it possible to install the router on it???

[JoChan]

On 11/8/2022 at 9:03 PM, SeanLMG said:

 

Save 59% on a year of Bitdefender Premium Security at: https://lmg.gg/qMMV8

 

Did you know that you can build your own router that can perform better than off-the-shelf products? It’s easier than you might think to build your own opnSense or pfSense router, and we’ll show you how!

 

 

[JoChan]

Hi , does any body know how to make you own router if your Internet provider is working with fiber modem/router?

Can you just install a SPF card in the pc and will I work ?