Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Because Everyone (Still) Needs a Router

codinghorror.com

Post reply on HN
Print view View original

121–130 of 137 posts

Re: Because Everyone (Still) Needs a Router

#121

Post by StavrosK »

tnorthcutt wrote: post #47
dugmartin wrote: post #5

As I commented here: http://news.ycombinator.com/item?id=4082351 The part of Tomato I like best is its simple DNS interface that lets you have one hosts file that is shared across all the machines connected to it. This is where I define my local dev domains so that I can test across devices that don't allow local host file changes (namely non-jailbroken iOS devices).

Whoa, that's pretty awesome. Does anyone know if DD-WRT can do this as well?

As paxswill said, it can, through dnsmasq. I've done it, it worked great.

Re: Because Everyone (Still) Needs a Router

#122

Post by icebraining »

speleding wrote: post #114
jwr wrote: post #111

Just as a data point on the Tomato firmware. I've been running it for about 2 years now, on two different hardware devices. It hangs every once in a while, the intervals range from several days to several weeks. I determined the culprit is QoS: if I disable all QoS, the device will run fine for months. Enabling it shortens the uptime to days or weeks at most. Since I know this happens reproducibly on two different de…

QoS in your router is only interesting if bandwidth is constrained in your last mile, otherwise it just adds complexity. I would think that QoS in home routers is only a transitional technology for the next few years while not everyone has a large enough pipe yet. Can you imagine your electricity or water outlet at home being constrained like bandwidth is? Those issues were solved a century ago, they'll get fixed for…

I can imagine, since my electricity outlet at home is constrained like bandwidth. We pay a fixed monthly fee that is proportional to the load capacity, so choosing a lower value and watching what I use saves me money.

In any case, I'm not sure if we'll ever have a large enough pipe for everyone; our current pipes are more than enough for our needs five years ago, but our needs grew, and will in the future.

Re: Because Everyone (Still) Needs a Router

#123

Post by dtaht2 »

krakensden wrote: post #101
sheraz wrote: post #44

Earlier quoted context omitted.

yeah -- I came across their site the other day. Would love to see some quantitative / qualitative numbers from that project.

They've done a bunch of work on reducing buffers and setting up AQM by default, and I've heard mumblings about improvements but can't find numbers. Closest thing I could find was this github repo: https://github.com/dtaht/deBloat/tree/master/test with lots of tests but no comparisons (stock OpenWRT vs ceroWRT would be interesting, for instance).

(dave taht here)

Given that core components of what we were trying to do (BQL and the codel and fq_codel qdiscs) have only just landed in the mainline linux kernels, doing comprehensive benchmarks would have been misleading.

Secondly - cerowrt's intent is a research project - everything that works we try to get into openwrt - so while the qos implementations differ, both are using fq_codel now.

It's my hope that tomato and gargoyle and dd-wrt - indeed as many router distros - adopt fq_codel or something like it as their underlying qdisc.

http://queue.acm.org/detail.cfm?id=2209336

thirdly - for the kinds of things we are trying to fix (bufferbloat, mesh networks, dns, ipv6) not a lot of good benchmarks exist.

Re: Because Everyone (Still) Needs a Router

#124

Post by dtaht2 »

krakensden wrote: post #101
sheraz wrote: post #44

Earlier quoted context omitted.

yeah -- I came across their site the other day. Would love to see some quantitative / qualitative numbers from that project.

They've done a bunch of work on reducing buffers and setting up AQM by default, and I've heard mumblings about improvements but can't find numbers. Closest thing I could find was this github repo: https://github.com/dtaht/deBloat/tree/master/test with lots of tests but no comparisons (stock OpenWRT vs ceroWRT would be interesting, for instance).

As for benchmarks, I have not taken the time to publish the results we currently get with fq_codel. However they are better in every respect than our previous attempt with sfqred and comparable to the qfq work we also did. It's not every day you go around seeing orders of magnitude improvements in anything...

http://www.teklibre.com/~d/bloat/pfifo_fast_vs_sfq_qfq_log.p...

Re: Because Everyone (Still) Needs a Router

#125

Post by dtaht2 »

rb12345 wrote: post #119
krakensden wrote: post #101

Earlier quoted context omitted.

They've done a bunch of work on reducing buffers and setting up AQM by default, and I've heard mumblings about improvements but can't find numbers. Closest thing I could find was this github repo: https://github.com/dtaht/deBloat/tree/master/test with lots of tests but no comparisons (stock OpenWRT vs ceroWRT would be interesting, for instance).

It's been a while since this was done, but as well as the AQM and Byte Queue Limit support, a lot of work was done on minimising unaligned access traps ( http://www.bufferbloat.net/issues/360 ). The Linux networking stack assumes that IP packets are word-aligned, but Ethernet headers are usually 14 bytes. The specific Atheros chipset used in the WNDR3700/WNDR3800 does not pad Ethernet packets and does not support mis…

I have not had the time to generalize these patches. I do note that alternate ideas for this problem landed in Linux 3.5 which are impossible to backport into Linux 3.3, so I have shelved the idea of working on generalization.

These patches were specific to the ar71xx hardware and unneeded on most other devices.

Re: Because Everyone (Still) Needs a Router

#126

Post by dtaht2 »

ShabbyDoo wrote: post #17

I posted this "Ask HN" awhile back about my desire to pay for a router-as-a-service: http://news.ycombinator.com/item?id=1160585 What surprised me was the number of responses along the lines of "you don't need this -- just install DD-WRT, OpenWRT, whatever and configure the foo,bar, and baz features as such." I'm likely in the 99th percentile of the US population in my ability to do such a thing, and I still have to…

the bufferbloat project's end goal is to get to where the default firmware on everything is good out of the box, and that involves trying to fix upstream for the kernel, tools, and distros, so vendors take it and run with it.

Re: Because Everyone (Still) Needs a Router

#127

Post by rb12345 »

jiggy2011 wrote: post #120
orbitingpluto wrote: post #80

Earlier quoted context omitted.

nmap is smart enough to determine what service is being offered on a given port and what OS is running. The above combined with the offending computer usually being a Mac and it's named something like "Bob Smith's Computer" is enough to common-sense narrow it down within a minute or two without having to use passive mode. (Also with the repeat offender I kept his MAC handy.) Then just initiate as many TCP connections…

Won't the program listening on the socket just drop the connection if it's getting stuff that doesn't make sense or correspond to a session that it is aware of? Or is the idea just to spam multiple connections?

It sounds more like connection flooding to me. TCP SYN flooding would also work, although modern IP stacks should cope reasonably well with that if syncookies are used.

That said ... considering these are Macs, they probably have mDNS open to IPv6 link-local traffic. It might make more sense to flood the offending machine with valid, unicasted mDNS packets. I don't know how well the Apple mDNS daemon copes with high traffic volume, but in my experience Avahi (on an Atom-powered netbook, admittedly) can regularly use over 50% CPU on a wireless network with an oversized local subnet. Hypothetically, deliberately flooding an OS X system with complex but valid mDNS announcements could have interesting results...

I should probably point out that Windows systems tend to be protected from this. Firstly, iTunes or Bonjour must be installed separately. Secondly, Windows Firewall tends to kick in with its "Public" profile, blocking inbound traffic by default. Thirdly, Windows machines tend to use mDNS over IPv4 multicast instead of IPv6 unless an auto-configured external address exists. This further reduces the traffic seen, although this last point is no protection from deliberate floods.

Re: Because Everyone (Still) Needs a Router

#128

Post by ajtaylor »

martinp wrote: post #31
btgeekboy wrote: post #7

No mention of Mikrotik? I picked up a RB750GL last week, and so far, it's everything DD-WRT/Tomato/etc wish they could be. It works as a basic plug-and-play router, but it's incredibly flexible beyond that. Check out http://wiki.mikrotik.com/wiki/Manual:RouterOS_features if you're interested. You can download the OS and run it in a VM if you want to give it a try before purchasing.

Another vote for Mikrotik. I bought a RB751G-2HnD myself about a month ago, because I needed a router with decent features and fast enough to handle a 90/10 Mbit connection. The feature list is very impressive and on par with professional-grade gear (Cisco etc.). Only downside is that it's not open source, but it's updated about every month or so. [1] [1] http://www.mikrotik.com/download/CHANGELOG_5

You've convinced me! I'll be picking up an RB751U this week. Now I need to return the piece of junk Linksys that stopped syncing to my DSL line this week. The best part? The MikroTik is cheaper but better!

Re: Because Everyone (Still) Needs a Router

#129

Post by nitid_name »

speleding wrote: post #114
jwr wrote: post #111

Just as a data point on the Tomato firmware. I've been running it for about 2 years now, on two different hardware devices. It hangs every once in a while, the intervals range from several days to several weeks. I determined the culprit is QoS: if I disable all QoS, the device will run fine for months. Enabling it shortens the uptime to days or weeks at most. Since I know this happens reproducibly on two different de…

QoS in your router is only interesting if bandwidth is constrained in your last mile, otherwise it just adds complexity. I would think that QoS in home routers is only a transitional technology for the next few years while not everyone has a large enough pipe yet. Can you imagine your electricity or water outlet at home being constrained like bandwidth is? Those issues were solved a century ago, they'll get fixed for…

I doubt water usage per capita has changed drastically in the last 20 years. Sure, there were drops from low flow toilets, low flow shower heads, and HE washing machines, and I'm sure there's some variability to usage based on seasonal variance from year to year, but the average US household uses about 350 gallons a day. I don't have access to historical data, but I'd wager that it hasn't changed more than 25% in the last decade or two.

Power has remained relatively constant at about 1.3MWh per capita. In the last thirty years, it has grown only 300KWh.

Internet bandwidth, on the other hand, went from 9.6-14.4kbit/s modems in fax machines in the 80s to 28.8-56.6kbit/s modems in the 90s to the megabit/s range in the 00s to the 15-100Mbps you can get in a residence today.

If the bandwidth issue is to get solved, the growth in consumption will need to peak.

Re: Because Everyone (Still) Needs a Router

#130

Post by vr000m »

I dont know what enabling QoS means on routers: is it traffic shaping or using QoS code points. Traffic shapping should be possible but I am not a strong believer of middleboxes tampering doing this (they probably dont have enough context to do it right). I'd rather have the applications or the browser or the kernel of the endpoint do this. For the latter this is kinda impossible because the QoS are ISP-specific and the application or middlebox should be aware of the them. Since there is no global solution or generic DSCP codepoints using them is not possible, however, there are some proposals in works to make generic marking possible.

The routers can enable WRED to mitigate some congestion problems, though it is not the silver bullet and works on a per-flow basis. Some new solutions to solve the buffer bloat problems is to install the experimental CeroWRT firmware in your routers (http://www.bufferbloat.net/projects/cerowrt), however, this is also a work in progress.

Post reply on HN