Router reliability musings
Having been asked a number of times in the recent weeks about routers and modems for both ADSL2+ and VDSL2, I figured it was time to revisit the subject.
I wrote about it some 2 1/2 years ago now, and unfortunately not a lot has changed. ISPs are still giving out cheap routers by and large. People are still having silly little issues with their routers. Connections still drop more than they need to (And for some reason people just ‘put up with it’ and presume its how things ought to be, presumably because they don’t know better). Sadly, there’s just as much confusion as there ever has been.
Thankfully its not too difficult to get a solid connection, and while some people may balk at the idea of having to purchase their own router / modem devices, a lot of people will find that the juice is worth the squeeze.
Lets start by addressing ISP-supplied devices. Not many people know, but I’ve been working for one of the smaller ISPs for a few months, and we dabbled in selling the same branded devices that the “big players” also offer.
I can also tell you a few things I know having worked as a consultant for a supplier of Dynalink / Netcomm devices.
Basically, there’s a reason why ISPs supply Thomson, D-Link, 2wire or Huawei devices, and it’s not because they’re awesome. It’s because they’re cheap.
Back around the year 2000 when New Zealand first received “Jetstream” broadband (ADSL1), the uptake was decent, but sort of slow. Most people were relatively used to paying $29-ish a month for unlimited dial-up, but a lot of people were still on cheaper plans. Add on to the fact that ADSL was going to double the cost of that, plus $100-200 for a router / modem and it hampered the uptake. So then Telecom had a brilliant idea: Lets give away routers and modems to help increase uptake!
Now back 12 years ago, that was a great idea. It helped people get online, people only had one PC, maybe two for the better part. PCs didn’t ship with anywhere near as much RAM, 512MB was a lot, so computers also only had a couple of apps open at a time. This in turn put less load on routers as well.
Why does this matter?
Well load on routers increases the heat. What increases the load on a router? Anything that keeps a connection open, or opens a lot of connections. Bittorrent is notorious for this. Skype was back in the day, I’m not sure because frankly I’ve not used it in a year or two now. SIP (VoIP) calling is rough though, it makes baby routers cry. I know, because I’ve both done VoIP for a living for a few years, as well as having watched the likes of Orcon fail with their “Genius” router. However, that’s a story for another day.
Anyway things like that all add load to a router, and many cheaper routers don’t survive under it because they get warm, this in turn affects your DSL connection which eventually drops.
Other routers have a limited amount of memory, which means that the memory that would be used to keep track of their “state table” fills up very quickly, and it kills off older connections because it knows its running out of memory and won’t be able to take any new connections if it doesn’t. This is how things like VoIP calls also end up dropping.
Then there’s just poorly written software for some routers. The likes of the Thomsons are notorious for DHCP and Encryption issues. For example, many of them will start to “throw a wobbly” after they’ve given out half a dozen DHCP leases and just not accept any other devices on to the network. Other stupid things that cheap routers are prone to doing is not playing nice with a larger number of devices while using WPA / WPA2, and forcing you to use WEP or no encryption. You shouldn’t have to use anything but WPA2 in this day in age.
So back to the story, basically it was a great idea that Telecom (Xtra) had back in the day to start giving away devices. However, in order to do-so, they had to give away basically the cheapest 1-port device they possibly could. This was the D-Link DSL-302g. It was riddled with security vulnerabilities over the years and was an all around “Cheap and nasty” device. These vulnerabilities included the device giving out the DSL username & password via the WAN interface. In otherwords if you scanned a specific TCP Port on all the ISPs DSL ranges (As I did a few times), found which devices responded, then ran a mass script to grab a certain file from them, it was easy enough to get somebodies account name and password. From there you could do everything including increase their data cap online, to shutting down their account altogether if you really wanted to.
It was good, it got ADSL uptake to skyrocket compared to what it would otherwise have been, and it was also very bad. I spoke with a friend in Telecoms Shared Services division a short while back who said they were still trying to pull the last of those crappy devices off their network.
Telecom also started to use 2wire devices. These particular devices were great because they were easy to mass-provision, and they’re also remotely configurable. The plan was that Telecom would be able to login remotely, have a look at what was going wrong, and fix it for you. The flaw with this theory is that it wasn’t the end-users 99% of the time that were causing the issues, it was either a fault with their wiring (Which Telecom almost always deduce without touching the router) or it was a faulty router giving the end-user issues. The theory behind it was great, but it just wasn’t terrible well thought out.
The 2wire routers didn’t last terribly long as Thomson came along riding in on a white pony with the TG585v7 (And subsequently the TG585v8 which is the same as the v7 but just with N-wireless, though only one antenna), saying “Look we can give you all that 2wire are giving you, but it’s going to be so much cheaper!”. How much cheaper? I’m unsure to be honest, but I know that it’s not difficult to pick up the VDSL2 Thomson (Technicolor TG789vn) devices for somewhere in the vicinity of $180 without even having to place a massive order. Given that ADSL2+ devices are generally in the vicinity of 25-30% the total price of VDSL2, then add in the face that you’re Telecom and you’re ordering a few thousand, it’d be even cheaper, I’d suggest they’re getting them from somewhere close to $40-50 a pop. I know that some of the more ‘expensive’ (Read: reliable) brands go for ~$50 a pop, so a conservative $50 is probably pretty close.
Unfortunately I’ve not been able to get an accurate figure, but if you’re reading this and you’re able to tip me off, then I’m all ears. I do however know that Technicolor will only customize units if you’re committing to order more than 50,000 units in a 12-month period, so basically Telecom would have committed to getting in a truckload of them before they realized the “quality” of them.
So anyway, Telecom are giving away these $50 routers for free. They’re able to do wifi, they have 4 ethernet ports on the back of them, so what’s the issue?
Well, the issue is they suck.
Don’t get me wrong, I know people who’ve been using them for months without missing a beat, however that’s usually just one, maybe two devices attached to them.
These routers were heralded on TV ads and everything as being the device that would “break down all walls” and allow you to “Surf in one room while streaming video in another”, for example (I can’t remember the exact wording of the TV ads, it was a while back). You would presume from the way they were portrayed on TV that they were the *ideal* family router that would work happily with multiple devices.
Unfortunately by the time you load on your printer, cellphone, a couple of laptops and maybe a desktop, they’re pretty much rendered useless.
What breaks? DHCP stops giving out leases, wireless has to be turned back to WEP encryption, or encryption turned off entirely. They overheat because they’re slow old devices with not much RAM, and the firmware is badly written. Don’t even get me started on the Web UI, that interface is absolutely freaking terrible! For the record, they have a 240Mhz ARM CPU (not too bad, should be sufficient), 4MB Flash (The bare minimum to be useful, but not the end of the world) and 16MB RAM (The bare minimum). The RAM is where the issue mostly lies I believe, combined with terrible firmware.
To give you a good comparison, my Draytek DV2750 has 64MB RAM and (I believe) a 266Mhz ARMv9 CPU. The TP-Link WR740N that Gargoyle recommends has 32MB (A ‘good & ample’ amount by todays standards).
Also, have you ever tried turning on content filtering, on one of those Thomson devices? That thing will fall over faster than you can blink, your throughput will absolutely *crawl* to a halt! What’s the point in having a feature like that? Can you imagine it’s like having air conditioning on your car, but as soon as you turn it on your car can’t go faster than 15kmph. That’s just retarded!
Now here’s what I’m finding difficult:
Device vendors change things, but when they do-so they don’t change the model number!
For example:
The TP-Link TL-WR740N that Gargoyle recommends and uses on their website is excellent! It’s well supported by OpenWRT (The base for Gargoyle) and the hardware is solid all around. The original firmware on the TP-Link is mediocre, but it does the job and it’s relatively solid.
Then, they go and change things with Hardware Version 4.20, quite drastically in fact. Suddenly OpenWRT no longer works (Support is added now, to the experimental builds), Gargoyle firmware also isn’t supported… Yet, if I’d told you to go and get the TL-WR740N and you’d picked one of these up, you’d consider me a liar and begin to doubt my technical prowess.
Same thing has happened with the NetComm NB6Plus4Wn. After a few people on PressF1 reporting issues, I decided to look in to the issue.
It doesn’t take much to look back through my blog history and see that I’ve been recommending them for a while now. I’ve literally bought and on-sold close to 3-dozen of them myself just with family and friends, let alone the *countless* people online who’ve asked for my suggestion and gone and purchased them.
You see, I’m pretty much the reason why PB Technologies used to say this on their website about the router:
“Troubles with your current router failing? Suffering from disconnections? Try this router, guaranteed to fix your issues or your money back!”
I spoke with PBTech back in the day (Around 2+ years ago now I think?) and told them about this router. They tried it themselves, got excellent feedback, and thought it was a great selling point for what was otherwise being overlooked as one of the more expensive routers. People had been opting for cheaper D-Links and Belkins as upgrades or replacements from their ISP-supplied devices and many were still having issues. Same even for some of the Linksys. In fact I’d been having a great time with my AM300’s and my WRT54GL’s, so I went out and bought 6 of the WAG160N’s for myself, my family, and a couple of friends.
That was a mistake, they were terrible, and I noticed issues within about 48 hours! Thankfully, PBTech were pretty good and took them all back.
But back to the NetComm NB6Plus4Wn devices. I’ve got in touch with PBTech now and see that they’ve cottoned on around the end of last year to the fact there’s a Hardware Revision 2, and they’re having some pretty major issues. Right now it’s expected that it’s a firmware issue that NetComm are looking in to, but that’s not really the point. The point is the hardware / firmware has changed, and the hardware is no longer as reliable as it was.
This means that if you come across my old blog post, or any of my older posts on PressF1, you’ll see “NB6Plus4Wn is great, go for it”, but there’s obviously no mention of “This is only Hardware rev1.0”, because back then that’s all there was. Now, we’ve got a newer version ofthe “same device” and it’s not working for people. I’ve got reports of it overheating, DHCP issues, port-forwarding problems, the works!
What was once known the *the* device that you would turn to, is now potentially even worse than the cheap junk that ISPs give away!
So what can you do?
Well unfortunately even the “best of the best” aren’t without their share of issues. It turns out even the Draytek DV120, which is quite possibly the most solid ADSL2+ device ever, still had some issues with UDP packets and needed a firmware upgrade. Same for the Draytek DV2750 (For VDSL2), you *have* to turn off the “Hardware NAT” feature, it just doesn’t work right at all.
Is it just me or can *DSL device vendors not seem to get things right?
I mean, surely it can’t be *that* hard, you’ve got vendors churning out dozens of different devices, why not just take the time to take one device, make it reliable, and keep it that way?
To be honest I don’t know what device I should be recommending at this stage. Without a doubt the Draytek stuff is top-notch, but it’s still not without its imperfections. I’m still using Draytek in tandem with Gargoyle and Tomato devices at home and for my family / friends, it’s what I deploy when people ask me, so that’s probably a great place to start.
In terms of a reliable “all-in-one” device? Again Draytek do some great stuff, but it’s not the cheapest. My theory is this:
Cheap, reliable, features; pick two.
You see the Linksys AM300 is basically used as just a “bridge” from ADSL to some other device when I use them. They’re great, reliable, they’re cheap, but they have almost no features at all aside from this half-bridging support.
There’s the Draytek DV120 is also the same.
It’s a little different when you get to Access Points like the Linksys WRT54GL and the TP-Link TL-WR740N, because a lot of things are added with the custom firmware, however again, reliability can be hardware dependent too.
It’s mostly the “all-in-one” devices that have this issue. If they are reliable and feature-rich like the Draytek DV2750’s, you can easily spend $450+ on a single device, just to get you online for VDSL2.
The NB6Plus4Wn was cheap, was reliable, and had some semi-decent list of features (Though not so much). Unfortunately now it’s just fallen in to the “cheap” category because the reliability is out the window, and also half the features don’t actually work.
What would I suggest?
I would suggest you try a device, and if it’s not reliable, try another brand and make and model altogether. Don’t just go from the Linksys WAG160N to the WAG320N. Instead, jump ship entirely and try something such as the Netgear routers (Which I’ve been heard a lot of good things about lately).
Personally I’ve left ADSL2+ behind at the start of this year (2012), and wouldn’t want to ever go back again. Pricing is coming down significantly, and the higher upload speeds are great (The download is good too of course), which makes it difficult for me to actually make a recommendation.
That said, if it was me, I would still go for a small 1-port device such as the Draytek DV120, and bridge it (It does PPPoA -> PPPoE Passthru, instead of halfbridging like the Linksys AM300) to a larger gruntier device like the WR740N, running either Gargoyle or Tomato. Again though, check the hardware version of the WR740N before you buy! You don’t want to get Rev4.20 which is currently not supported by Gargoyle.
If you’re on VDSL2, then you can’t go past the Draytek DV2750!
Have you tested a Netgear device yourself that you can recommend?
What about some other brand? I’ve heard semi-decent results about some of the newer Belkin kit that always leaves me hopeful!
If this has helped you, leave a comment and say hi, or drop me some feedback, I’m always keen to hear from readers.
Chilling Silence, thanks – terrific effort to write all this up!! I left a question on your older blog which is pretty much answered here. I have found both DV120 (DrayTek Vigor120) and TP-LINK TL-WR740N at PB Tech, and have just asked them to ID whether it’s a v4 or other hardware. (See link http://www.tp-link.com/en/Support/download/?model=TL-WR740N).
My remaining questions are, if I buy these two, what are the steps then? Sounds like…
– upgrade DV120 firmware to latest
– put tomato or gargoyle on TL-WR740N
My aim is to be able to monitor the household usage, see what devices are using it, and perhaps throttle it. Knowing who’s using what is the priority, but being able to set caps and (even better) time windows by device, would be ideal.
So…I’m a bit hazy on how to set it up;
– does the DV120 just need DSL authentication (my ID an PW as I have static IP) and then to be put into PPPoE to PPPoA
Bridging mode?
– Is everything else then done on the WR740N? What LAN settings go into this device? I presume a static address on same internal net as everything else (10.1.1.0/24)
I have a household with a small business server, with its own DHCP and DNS. I have 5 kids, so there’s a Snow Leopard Mac, 3 x Windows 7 and 1 x Windows XP PCs, Droid phone, two iPhones and 3 iPods doing wifi.
I have a single point of entry for the DSL in the house.
I use an old Telecom D-Link DSL-G604T for that, with the wireless turned off.
It goes into a 5-port 1000 Gbps Linksys EG005W-AU switch.
Out of that switch is a Linksys WAP54G wireless serving the house (G wireless).
A 50 m ethernet run takes the network over to a separate office/garage.
In there I have an 8-port 1000 Gbps Linksys EG008W-AU switch.
Out of it there’s another (yep, ex-Telecom!) D-Link DSL-G604T wireless serving that area (same SSID, different channel).
Then there’s a few PCs and the SB Server there on ethernet.
– DHCP and DNS are done by the SBS
– DHCP uses .100-.199 range
– ADSL Router is .1 fixed
– SB Server is .20 fixed
– Wireless routers are .19 and .18 fixed
I have replaced Xtra’s DNS with Google 8.8.8.8 and 8.8.4.4, at the router.
So, I imagine I would…
– replace the DSL-G604T with the DV120, in PPPoE to PPPoA Bridging mode.
– Then add TL-WR740N with tomato/gargoyle, using a fixed IP (logically, 10.1.1.2).
– Probably run wireless for main house out of that device, and retire the Linksys WAP54G or move it to office to replace the other DSL-G604T doing wireless there.
Everything else from there would be the same. This should give me a single point for monitoring and managing bandwidth.
How’s that sound?
Hi snoi,
Sounds like you’re on track, however I’d personally opt for Gargoyle on the WR740N rather than Tomato, because you can do per-device (It uses the MAC Address) bandwidth restrictions and accounting. You can set daily / weekly / monthly caps for each independent device, or apply a blanket “Everybody gets 2GB a week” rule. Once they hit the limit, you can throttle them back to XYZ speed determined by you, or you can block access entirely.
Just make sure that you force Gargoyle to ignore any device that it hasn’t given out an IP Address for, otherwise somebody in your household could simply statically assign themselves a different IP Address and get a “new 2GB” for example. By doing this you enforce that Gargoyle will shut them out if they try and circumvent it.
The DV120 doesn’t do any authentication at all, it simply acts as a “bridge” between PPPoA (The way you get your internet) and PPPoE (How your Gargoyle / TP-Link router will authenticate). The DV120 basically becomes a “dumb” device then just relaying everything between your Gargoyle router and your ISP. The Gargoyle Router does all the hard work then, and also the authentication, so you specify your “WAN” connection type in Gargoyle as PPPoE. The WAN port of the TP-Link Gargoyle router then goes in to the LAN port on the DV120 which is set to bridge your ADSL2+ connection. Hope that clears it up.
The best part about this is it’s going to work both on Windows, with your server, with your Android / iPhones / iPod Touches etc. and you can also make your small business server exempt from any throttling back or limits, as the last thing you want is to mess with that.
I would personally opt to have Gargoyle give out addresses via DHCP and the SBS box still do DNS. Gargoyle can intercept DNS and relay it off to XYZ other device if you specify it, there’s just a tickbox, so you could specify your SBS box as primary and then Google DNS as secondary for example. This still allows LAN-local DNS stuff to work as applicable.
I’d be inclined to throw Gargoyle on your WAP54G and use it as a Wifi-repeater. Means you can leave it standalone to cover a greater distance and potentially eliminate the DSL-604T. One less device to worry about.
But yeah basically what you should probably do is get it setup, leave everything else in-tact, and simply replace the current DSL modem you’ve got. Make sure it works nicely, and potentially even give it a different SSID while you muck around with it. This way if it takes, say longer than half an hour or an hour, and you need to get back online in a hurry, you can just throw your older router back in and have things keep going and tackle it the next evening.
Sounds like it’s most certainly doable, shouldn’t be too difficult either, and no doubt it’ll be a fun little project. It’s a very similar setup to what I’ve got going at my parents place, I’m from a family of 6 kids and we all drop back in at home from time to time (4 still living there) so traffic accounting is incredibly useful.
My advice: Give everybody a 1GB limit for the first week or two. See who comes running the fastest, keeping in mind that they can alternate between their laptops and iPods, and each will have an independent 1GB cap, so they’re not completely cut off from the world. You’ll see very quickly who’s leaving uTorrent running at nights 😉
Post back later and let me know how you go.
By the way, I got my TP-Link from PBTech, they’ve only got the v4.20 as far as I’m aware. Get them to pull it out of the box, in fact if you go in-store they have a security guard who will open the box for you (They’ll shoot you if you do it yourself) and they’ll check the version. It’s underneath the device on the top-right hand side of the sticker, it’ll say Rev 4.20. If it does, don’t buy it coz it ain’t supported, and you should try for another model. Speaking from experience here, I’m stuck with OpenWRT on mine at the moment…
Good luck!
[…] on from my earlier post last month on “Router reliability musings“, I recently did some additional “testing”, and figured now was a good time to […]
Thanks – your posts are really useful. I just read your follow up post at http://www.c2s.co.nz/blog/more-router-reliability-musings/#more-921.
I am really looking forward to this project and getting bandwidth under control, but do you think I can find a Gargoyle-supported router to buy? This has stalled me for a couple of months now. I’d like to find a 32 MB RAM / 8 MB FLASH modem that will work with Gargoyle. I have accounts at PB Technologies and Acquire, but just can’t find a supported modem. I contacted PB via their web page to ask what version their TP-LINK TL-WR740N is (http://pbtech.co.nz/index.php?z=p&p=NETTPL0742) but got no response and got too busy to persist. That modem also says “(no ADSL built-in)” so what’s up with that? I guess it’s not going to work here.
Is there a NZ supplier I can get a suitable Gargoyle modem from?? I’ve been trying to work backwards from what I find in stock, but everything turns out a dead end so far. pricespy.co.nz are great, but none of the suppliers so far are folks I can get to or have check hardware versions for me. Arrrgh!
Yeah the 32MB routers are ideal. Everthing else really should run an older version of Gargoyle, I believe after 1.3 they changed some stuff and made it basically *require* it.
In a nutshell, anything supported by Gargoyle or Tomato won’t have ADSL on it. That’s the basic rule. This is why you need either a Linksys AM300 in half-bridge, or a Draytek DV120 in PPPoA -> PPPoE Passthru mode. You can tack the Tomato / Gargoyle router behind any other DSL Modem / Router, but it means you’re then going to suffer double-NAT and things which can be a pain in the ass, but still fine if you don’t need to do any port-forwarding.
You could try getting in touch with GoWifi, they’re pretty good there! I’d be surprised if half their stuff didn’t support it 🙂