Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Weaver]

Many users will know all this as I have been around for over 9 years so to them sincere apologies, but there are a few new users around [welcome!] who don't know about me.

I am a retired C and machine code programmer, packed that in 20 years ago, made a pittance as a consultant for a few years until ill health forced me to give up work.

For the past 7 years nearly I have been semi confined to bed although I can make it as far as the bathroom next door. I have various sorts of pain and also chronic fatigue syndrome which is just a posh way of saying that I feel incredibly overwhelmingly tired some of the time. I am on a lot of extremely powerful drugs for pain which mean that I am pretty out of it a lot of the time.

I have big problems with concentration and memory.

I have three extremely long IP-bonded ADSL2 lines with the ISP AA (aa.net.uk). I love my IPv6 and everything also has a global routable public static IPv4 address too as I have a decent sized routable IPv4 address block from AA which I got (they give IPv4 addresses for free!) before a good while ago before IPv4 really started to run out. I use a Firebrick FB2700 router and am currently switching to 3 * ZyXel VMG 1312-B10A modems instead of the DLink DSL-320B-Z1 modems I have been using for about five years. I get just over 3 Mbps sync rate downstream from the new ZyXel
modem, reality is 88% of that. The other modems are set to be replaced but give me 2700-2800kbps sync rates. I get a combined real 1.20-1.30 Mbps IP usptream and the Firebrick router combines the upstream pipes into one efficiently load-spreading all IP packets. Single TCP connections in both directions go at triple speed; I don't need three conversations/activities/users to see the benefits of triple line goodness, it works on a single straight download without even multiple TCP connections in use.

I use an iPad Pro (large screen model) solely, for absolutely everything. I write code in the Textastic editor.

That's about it. Might explain some of my oddities. 

[Deathstar]

We are all odd

[johnson]

Good to get to know you Weaver!

I'm sorry if I have seemed insensitive or rude in my posts as I guess I'm probably one of the new users you are referring to.

I had no idea line bonding worked as well as that, I always assumed it would be bottlenecked to a single lines speed for stateful things like TCP connections, with the real benefits only coming from aggregating multiple streams. I cant lie, I used to think a lot about combining many *totally legally accessed* wifi networks from the local area into a monster torrenting connection back when everyone used WEP.

[Weaver]

There are quite a few new users around, and some have been very helpful to me.

I thought a recap might be useful, because when some stuff is well known for years then it means people who are latecomers are just rightly mystified.

The AA (downstream) and Firebrick router (upstream) line bonding works extremely well, from the evidence of various measurements. My downstream is probably handled by one of AA’s own big Firebricks, an FB6xxx-series router, at their end.

I am told by AA the ISP that one should expect reduced efficiency when you get more than four lines, but some people do have rather more than four lines in fact, so I believe.

You can have a lot of lines on a small Firebrick by using VLANs for multiplexing multiple modems when you have run out of free ethernet ports on the Brick to plug more PPPoEoE modems into, that's how you cope with more than three.

My router has a single IPv4 WAN address facing towards the internet that is used by all the PPPoE interfaces. AA split traffic according to changing per-line data rates reported to them dynamically by BT or TalkTalk Wholesale systems.

They detect when a line goes down because they are PPP-LCP-pinging the router every second or so and if it is really down then the traffic goes to the other links and this happens incredibly quickly, within a second or so. They can also redirect to 4G with this link failure detection mechanism ('CQM' Constant Quality Monitoring). If any line goes down, I get tweets and SMS and email instantly.

The upstream is a bit less convenient because there is no automatic speed detection for upstream per-link rates. You need to tell the brick what rate to run each line at so it does the correct rate split.

I have written a tool to reconfigure my own Brick with new upstream rates remotely but you have to know what the rates should be to give the values to the tool. I am going to start writing something, a part 2, to discover the current upstream rates remotely and link that to the remote router config module already written, and the final stage three might be to make a continuously running monitor that automates the whole thing with regular rate queries and runs all the time. But that would mean moving the whole thing to a continuously available server box as the current stuff runs on an iPad and that is not necessarily always up - even though I do have a spare iPad iPads go to sleep.

[renluop]

We are all odd

Indeed

“All the world is queer save thee and me, and even thou art a little queer.”

― Robert Owen