For those who are curious, there's a bunch of reasons that contribute to why consumer wifi routers sucks ass:
1. Wifi routers are very complicated. You need smart people at all levels of the stack to build a modern wifi router. You need people who know 802.1ac like the back side of their hand, people who know how to set up and deploy linux environments, embedded engineers for debugging OEM driver issues, networking gurus for handling the voodoo in levels 2-4 of the OSI stack, application people for rolling the user interface, cloud people for cloud support. Normally this isn't a big deal, if it wasnt for the next point
2. The profit margins on consumer routers are complete trash. Even if you're one of the big boys with double digit market share, you're going to have a very hard time keeping a decent engineering team staffed and your marketing team staffed at the same time while still breaking even.
3. Consumer router sales are SKU driven. There are dozens of price and performance points you have to hit to meet the demands of the consumer market. You cannot make be profitable with less than 10 actively selling SKUs. Every time you release a new SKU, it's a new opportunity for marketing to try to sell the device to brick and mortar stores that they're trying to expand into. If you aren't releasing 5-6 SKUs each year, you're going to have a very hard time keeping your router on store shelves.
4. Since sales are SKU driven and your engineering team is probably under funded, you have the exciting problem of maintenance releases. If your company has 50 supported SKUs and you find a non-driver issue in one of them, the chances are that it affects 10 other SKUs as well, if not all of your other SKUs. Pushing that maintenance firmware to 50 SKUs could easily take 6 months of combined QA or Firmware development time. As far as your marketing department is concerned, all that time you spend on maintenance releases is time that isn't spent on making new SKUs with exciting new features.
My experience is that most consumer routers are actually developed and maintained by teams at companies you've never heard of overseas. This is partly as a response to the engineering challenges you've outlined.
Taiwan is a popular choice, due to the combination of proximity to manufacturing in China and an English friendly business culture.
Every so often the big brands go shopping to build a new product and they select something that meets their needs from the white-label products available.
The product then gets customised for them. What the big brands contribute is industrial design of the case, some qa, packaging, ui styling, a distribution channel and consumer trust in their name.
Which leads to funny scenes like watching engineers at a white label do board relayouts and emi retesting because the industrial designer wanted the antennas to sprout out of a different bit of the case. Because it looks more badass that way...
This also leads to the phenomenon where say, v4.15 of a router was cool, but 4.16 sucks - because internally its a totally different chipset, with a different os (e.g. vxworks vs Linux). Ref e.g. WRT54...
The reason this happens is not because the brand name rewrote anything - they just selected a different white label partner who customised a totally different product to look the same (probably because they promised to deliver cheaper for same SKU). This then gets sold under the old, popular model name.
This is why I was happy with my Apple products for years. No matter what kind of third-party thing I'd buy it would end up being a piece of trash after a little while. Apple Airports tended to just keep working for my family.
Too bad the new ones are ugly as hell. Oh well, my older model still works.
And these days so many people get it through their cable provider (or DSL, whatever) who have a marked interest in not generating support calls means you may get decent equipment.
> And these days so many people get it through their cable provider (or DSL, whatever) who have a marked interest in not generating support calls means you may get decent equipment.
You would think that, unfortunately it isn't true. Usually those ISP provided combination routers are completely outsourced - as in their hardware and firmware are both made by a chinese OEM overseas.
Believe me when I say there is nothing worse than the firmware on those chinese routers. They'll grab a bunch of open source software, reskin it, plug the chipset's reference driver into the OSS, then do a basic sanity test and ship it. You get what you pay for, and the ISPs want to pay as little as humanly possible.
I don't know what they do about the support. I remember one time we brought in an ISP's combination modem/router for performance tests against our device lineup, and it would crash if you ever tried to pull more than 40mbit/s through its ethernet ports.
Having worked at an ISP, the problem is that pretty much all the routers/modems at price points consumers will accept (because an expensive router, even if leased, means a higher monthly bill) suck.
They all work really well in the testing lab, where there's 50m of cat5 between the DSL modem and the DSLAM on the floor below. But then you find out once it's deployed that they have big issues with certain kinds of interference, so for a small segment of your userbase they have more dropouts than other routers. Since the company has already bought tens of thousands of them at this time, it's not really practical to just recall all of them and buy new ones. So the ISP files a bug with the upstream vendor. After some months of pushing, the vendor gives the ISP a special firmware build that's supposed to fix the problem. The fix never gets into the official firmware line (presumably because it has other downsides that the mfg doesn't want to talk about, maybe it's more conservative in line measurements etc), so now the ISP is locked into running this custom firmware version. Oh and that one probably has its own bugs that the older one didn't.
I saw this happen so many times, with routers from multiple vendors over my years doing this kind of thing. I don't run consumer networking equipment in my home anymore.
Fully agree with the comment about the quality of Chinese firmware. I spent the best part of a decade working for a UK company producing reference designs for mobile chipsets, which we sold to many Asian clients, and I lost count of the times I had to travel to the Far East to sort out the mess they'd made of the software.
Their philosophy seemed to be to throw as many inexperienced graduates at the problem as they could afford and eventually it would somehow work - as though an infinite number of engineers would eventually produce the works of Shakespeare!
I've never understood Shakespeare to have much renown as a software engineer, though I suppose I've never understood him to be a particularly poor one either.
I'm aware of the theorem, but I was being willfully ignorant to achieve a primary goal of making a joke. The secondary goal was to point out that the engineers the Chinese are hiring are probably better at programming than Shakespeare was. Regardless, the tone of my comment was in line with that of the comment I was responding to -- surely anybody hiring engineers would not judge the venture a success if they found the engineers had reproduced a number of plays from a particular Victorian playwright.
I switched cable internet providers because the original one (a large incumbent here in Toronto that rhymes with Ogers) wanted to force me onto their combo wifi/modem (the other large incumbent, rhyming with 'Ell, has the same constraint). I absolutely refuse to use an integrated device from the carrier at this point because I have experienced the crap that they use previously. Give me a piece of equipment that does one thing well, and let me pick the most reliable in-house networking equipment myself.
Standard procedure here is to ignore the device's onboard wifi and plug in your own access point. Don't tell them, or they may whine about it being unsupported.
You don't have to use the integrated access point just because it's there.
Sure, but you still wind up dealing with the craptastic device. The router I currently use which is forced upon me by my ISP will hang for a few seconds if it encounters a DLNA packet. That means several seconds of packet loss and some TCP connection being closed.
I have wasted so many hours on the phone with their support without getting anywhere. They wouldnt even acknowledge there is a problem until I showed them graphs. Then they sent an "engineer" over who also knew fuck all and accomplished nothing. He then seemed to have reported back there was nothing wrong and now they ignore me.
I should switch ISPs but the performance of the network compared to the price is excellent. Just this shitbox of a router.
At least in Germany we had the so called "Routerzwang". Some ISPs allowed third party equipment, but many didn't. You'd get a modem / router / AP combo device and the ISP had complete control over the firmware. Since there was no competition, most ISPs didn't care about the quality of their software or even security updates.
Thankfully in the UK all the (cheap) ADSL suppliers I've used have provided a modem+router+AP in which you can just read the access credentials off the config screen. No need to spoof MAC or anything when setting up a new 'router' just use the right credentials.
Now I'm wondering why, whether it's legislation or the way our ISPs relate to the infrastructure.
Primarily because our phoneline also comes from the same ISP which has to use their shitbox. They don't give you the credentials for that, you get a pretty configured router. It's possible to root the box, steal all necessary credentials and clone the mac but that is a huge hassle and obviously not supported.
I had similar issues, and I complained enough that my ISP gave me a second, just plain stupid modem and a splitter, so now I have their fancy phone/modem/router/wifi one just for the phone, and a simple modem only box for my actual internet. Works like a charm now!
Just keep in mind, their "first tier" of customer support probably can't authorize that kind of thing, and probably won't even know what the heck you are wanting. I complained about the crappy slow internet and bad ping until I got to their actual tech guys, and told him the older modem did better, and he was like "Well, why don't you use one of them?", I said "Need the phone", and he told me to just put a splitter on it and use both, he'd set it all up and have the local office put one aside for me. I only even pay rent on the one modem, so it's actually an amazingly good setup, other then meaning I've got the start of my own "commercial grade" networking setup attached to a wall in the basement, with the modems, the router, a switch, and a server.
In the netherlands providers has linked the wifi enable settings in their modem/router/accesspoints to their online account and if you disable the wifi you won't get access to the free networks they deployed everywhere.
This is mainly because those networks are deployed by adding an second SSID to the router with WPA enterprise auth. The solution many people choose is putting the router in the basement with a lot of aluminium foil around it.
I don't see why this is a problem. You want to free load on other people's connections but not share your own? (In the US anyway, the free wifi you get for running that secondary SSID just comes from other people's personal connections).
This is why we can't have nice things. I hope they find a way to defeat people doing this, but my guess is it would be prohibitively expensive (war driving entire cities, etc).
The problem is that I don't have the bandwidth for it. If a single user connects to the free accesspoint and starts skype r a youtube video then I cannot use the internet anymore.
I had this problem with my Optimum router. I actually got better performance by throwing it in the closet and using a $30 dollar piece of crap instead (which has since been replaced). Sadly, it seems that everything in the consumer networking market is a trap.
That's unfortunate. I never really use the Wi-Fi on my Comcast box so I don't know if it was any good. I've had great experiences with people Google fiber boxes, but that's Google and isn't exactly a mass deployment so…
Admittedly I haven't tried the newest models, but my experience with Linksys hardware and either ddwrt or tomato was pretty bad. Run for a couple of days and then require a power cycle. Seemed like a hardware issue.
WRT1900ACS is pretty good with DD-WRT (uptime can be until next firmware update). Linksys also sell special set of antennas, which improve signal range (costly but can be worth it depending on your situation). I think they target it for network enthusiasts, rather than very wide market, so they put some effort into quality there.
Counterpoint: I run dd-wrt on a TPLINK el cheapo something at the office, to connect all the boxen and let me SSH into each of them (port forwarding 19 to 22 on box1, 20 to 22 on box2 etc). It Just Works, never had to reboot it in four years now. (WiFi is off though.)
dd-wrt has gotten pretty bad. They don't publish new releases anymore you just pick a daily build from one of the random people who publish builds and hope it doesn't brick your router. Learned this after bricking my router :(
I think their main developer (Brainslayer) publishes his builds periodically. So those aren't random people. But in general it's not really a fully open development project. I.e. it's not governed by any community.
OpenWRT is better in this sense, but even there some people split into recent LEDE project: https://www.lede-project.org
Regarding bricking. Linksys WRT1900ACS is designed with that in mind. It has two partitions, and you always flash updates to another one, so if something goes wrong, you can easily switch to the other partition which will hold previous installation, using special on / off sequence. It's a neat idea, and more routers should follow it.
It should be noted that Ubiquiti APs have no router functionality, they're just plain access points. Which is the way I like it, but might not be appropriate for many use cases.
I bought myself an Ubiquiti Security Gateway and the AC Pro access point. Works really well and the total cost wasn't all that much more than, for example, an Airport Extreme…
Yeah, it disables all the hardware offloading unfortunately. But if you really need it because of a slow internet connection, I guess 60Mbps is good enough.
i actually edited my post to include the thing about openwrt/ddwrt.
If you need stability, go with a high end consumer router and load OpenWRT/DDWRT on it. OpenWRT and friends are very stable, but don't often take advantage of advanced features like automatic channel selection, traffic prioritization, or beamforming.
If you absolutely need features, go with a ubiquiti AP. My only issue with them is their hardware leaves a lot to be desired, and is rarely powerful enough to cover a full home, this isn't necessarily ubiquiti's fault though.
Business class APs expect you to set up multiple APs in the office, so their hardware is typically a lot less powerful than what you'd see in a consumer wireless router. A ubiquiti AP will gladly cover your 1 bedroom apartment, but don't expect it to cover your 3 bedroom home
The difference between your and GPs experience probably comes down to concrete vs. wood housing. WiFi, like most radio signals, has really poor propagation through concrete.
I ended up just keeping my ISP router as is, and putting a DD-WRT router behind it. I suspect the double layer of NAT isn't great, but I've been doing it for years without issue.
DD-WRT has been pretty solid for that, even in non-ideal network setups (currently have it connected to the ISP router as a 2.4GHz client, then acting as a 5GHZ AP for my other devices).
Personally I really like the new Draytek models (like the 2925 and 2860 series). Lots of nice features like central access point management (Draytek APs only) and central VPN management to configure remote Draytek routers are great features on top of fail-over/load balanced multi- WAN and even a model with built-in LTE modem as a WAN link. They're intended for always-on/highly reliable situations. You can even pair 2 units into an HA cluster.
They're very popular in Europe but oddly never seem to have had the same impact in North America, although you can find them if you look around a bit.
They did do a linux based variant at one point but switched back to their custom rolled OS a couple of years back. Not sure of the story there or if it's possible to do a custom firmware. There was something on google code at one point.
pfSense running on an old laptop or other lightweight system, along with a VLAN-capable switch and a Ubiquiti or Cisco WAP371 wifi point. Bridge pfSense to your modem, and you have a full stack. You can run your pfSense box on a stick via VLAN-ing, so you only need one interface.
Just be aware that if you're using any flavor of Comcast in the US (Business or Consumer-class), "Bridging" your cable modem will make it cease to work and require a factory reset. Anecdata from me, from sites in Boston, Seattle, and Portland.
Personally I prefer at least desktop hardware for my pfSense boxes, with a physically separate WAN and LAN port. Maybe that's just because I'm bad at VLANs.
I find most consumer-grade Wifi routers make more than adequate ACCESS POINTS, but they suck at being an actual ROUTER. My TM-AC1900 (rebadged RT-68U with a T-Mobile logo on it) would need to be restarted daily, after I finished setting up my homelab two months ago I put a Ubiquiti EdgeRouter X in front of everything and just connected the old router to my switch and put it in AP mode, I haven't had my network die out on me and require reboots in months (with the exception of my cable modem going out).
The problem with Omnia is that it's a niche product that is very unlikely to ever sell outside of its niche.
Also, it's being made by people who haven't made a router before. 99% of the issues you'll ever have with your wifi router are driver issues. Omnia is using OpenWRT for their base, however they'll still have to make the chipset driver work with OpenWRT, and at a reasonable speed.
Even if they get the driver working, it's unlikely that the device will be running at speeds comparable to other market competitors. It takes lots of unpleasant driver hacks, custom networking stacks, and prioritization magic to hit the benchmark numbers that Asus, Netgear, and Linksys push out.
I'm not saying you should write off the Omnia completely, I just recommend sitting back and waiting for it to be released/benchmarked before purchasing it.
What about AVM products such as a Fritzbox (https://en.wikipedia.org/wiki/Fritz!Box) ? Primarily a german product but also available internationally (e.g. Australia)
The Fritz!box is the only router I recommend to friends. I've run into few problems with them, except persistent issues with the web interface responsiveness. The only real problem is that they are relatively expensive. It's hard to convince people to spend 4x the amount because it's "better".
I think most non-geeks don't realise just how bad their routers are making their web experience.
Well, that depends on your perspective I guess. For most Aussies stuck on ADSL (real world max speed of 20mbps download, on a good day, close to the telephone exchange), their WiFi performance is rarely the rate-limiting factor.
I bought a Ubiquity a few months ago, and found out that it couldn't communicate over wireless with a Raspberry Pi's USB-WiFi dongle (with drivers in mainline kernel) sitting 3 meters away. Before I bought it, the same dongle was acting just fine as access point for the whole home network of 3-4 phones/tablets and 1-2 computers; it could easily handle browsing, updating Android/iOS apps, and YouTube video streaming. And it still is doing that just fine after I sadly had to return the Ubiquity!
Now of course this is just my own anecdote. But it remains a fact that WiFi setup can be incredibly hit-or-miss, and there's no reason in principle why two devices 3 meters away should have occasional 2-3 seconds packet delays.
It has nothing to do with the firmware it has to do with the fact that they haven't delivered a single working router to a customer yet. You don't seriously believe that buying something through a crowd funding campaign carries zero risk do you?
As a product, thats what I would have bought had it been available. Why have a weak router and a weak NAS if I can have a decent combined device with the same power consumption?
I think the problem is not with the actual IP routing part, as that is relatively straightforward, but with all the other value-added "features" that get crammed into the same and often underpowered hardware. At the least, a consumer WiFi router will contain an IP router, a NAT, and an AP, and it's the NAT that probably contributes a lot to the connectivity failures people are seeing that go away when the "router" is rebooted --- NAT tables filling up due to many half-open or lingering connections that didn't get closed properly. The aggressive nature of many applications' connection usage (e.g. "why does my router stop responding when I use torrents?") doesn't help either.
It's not just terrible NAT, although that's popular. I already do NAT on a PC server, so I run wifi routers in access point mode, and still need to do periodic reboots because the wireless interface stops working.
Is there any specific reason why consumer grade router companies (like D-Link) drop their own OS entirely and shift to a open-source alternative? That should bring down development costs somewhat I'm guessing and force them to make different boards compatible.
Not a hardware guy, but I've looked at D-Link firmware code in the past, and it is just utterly shoddy. Just making the hardware and making sure it is compatible with some router-os should be simpler. (Essentially what happens in the mobile/laptop industry with android/windows).
Maybe not exactly what you want, but Netgear has a fairly wide range of routers with open source firmware support and have a dedicated site [1] for that. I've had good luck with the previous 2 Netgear routers (WNR3500 and WNDR3700) and will probably buy another one again in the future.
It's pretty obvious that 2 and 3 are a big problem. I'd pay significantly more for a router that I can guarantee works reliably (cost doesn't seem to make a difference for consumer router reliability); find a way to advertise that and I'll love you forever. Make packet loss and latency on a variety of popular hardware / OS (with default settings) THE key metric.
Protip: if you are willing to pay more for reliability, buy Ubiquiti (or other enterprise class) hardware. It's actually not that much more expensive, and you get great features like PoE, design/color ment to be invisible, and great handover between multiple APs if you have a large house.
That's precisely what we're doing at eero. We have engineers focused on each part of the stack (wireless firmware + drivers, core OS, application layer, mobile, cloud, and data) and are pouring our efforts into a single sku. It's time that people had the network they deserve...especially since they run our homes.
People rarely recommend routers by their model number, if they recommend a router, they recommend the brand.
It's very hard for marketing guys to call Costco and say "hey, you remember that router we tried to sell you 6 months ago? well we just gave it a firmware update!", and close the sale
it's very easy for marketing to call Costco and say "hey we got this fantastic new router you guys might be interested in stocking", and close the sale
My BT HomeHub5 has been pretty unreliable - regular dropouts of Wifi from some Macs, similar to the DHCP thing that OP described. Then monthly reboots when everything stops. Performance is pitiful, but I blame that on the ADSL1 :)
I finally gave up on consumer routers, and bought one of these instead - the learning curve is steep, but it's overloaded with features and runs brilliantly.
My experience using draytek is that the documentation is non existent and the small amount that does exists makes the exotic bits of cisco documentation as clear as day.
1. Wifi routers are very complicated. You need smart people at all levels of the stack to build a modern wifi router. You need people who know 802.1ac like the back side of their hand, people who know how to set up and deploy linux environments, embedded engineers for debugging OEM driver issues, networking gurus for handling the voodoo in levels 2-4 of the OSI stack, application people for rolling the user interface, cloud people for cloud support. Normally this isn't a big deal, if it wasnt for the next point
2. The profit margins on consumer routers are complete trash. Even if you're one of the big boys with double digit market share, you're going to have a very hard time keeping a decent engineering team staffed and your marketing team staffed at the same time while still breaking even.
3. Consumer router sales are SKU driven. There are dozens of price and performance points you have to hit to meet the demands of the consumer market. You cannot make be profitable with less than 10 actively selling SKUs. Every time you release a new SKU, it's a new opportunity for marketing to try to sell the device to brick and mortar stores that they're trying to expand into. If you aren't releasing 5-6 SKUs each year, you're going to have a very hard time keeping your router on store shelves.
4. Since sales are SKU driven and your engineering team is probably under funded, you have the exciting problem of maintenance releases. If your company has 50 supported SKUs and you find a non-driver issue in one of them, the chances are that it affects 10 other SKUs as well, if not all of your other SKUs. Pushing that maintenance firmware to 50 SKUs could easily take 6 months of combined QA or Firmware development time. As far as your marketing department is concerned, all that time you spend on maintenance releases is time that isn't spent on making new SKUs with exciting new features.
tldr: get a business class wifi router