from the 640kB-ought-to-be-enough-for-anybody dept.
Internet outages and slowdowns spiked earlier this week, and more are likely on the way in coming weeks, as the internet grows too big for some network hardware to handle.
The issue arose when the number of routes on the internet temporarily jumped beyond 512,000 or 512K -- the maximum that some older networking gear can handle by default.
Cisco tweeted: "The global routing table has passed 512k routes. So what now? #IPv4 #IPv6 Read: http://t.co/gJOafqLGDk via @CiscoSP360"
The cause was a bug at U.S. internet service provider Verizon that dumped 15,000 new internet destinations onto the network for about 10 minutes, said Andree Toonk, founder and lead developer for BGPMon. "We basically got a small taste of what is possibly about to happen," added Toonk, whose company monitors internet routing for outages and security incidents. "Hopefully this is a wakeup call."
Network analysts such as Toonk estimate the number of routes in the internet -- currently hovering around 500,000 -- will permanently surpass 512,000 within a month. The hardware causing the problem is older routers made by San Jose, Calif.-based Cisco and still used by many smaller networks and regional internet providers, Toonk said.
(Score: 5, Funny) by Ken_g6 on Saturday August 16 2014, @09:17PM
(Or at least what people claim Bill Gates said. [wired.com]) They should have used 640k.
(Score: 2) by isostatic on Sunday August 17 2014, @06:44AM
With ip4 you could theroetically have 4 billion entries - every individual ip could have its own route, up to 256 hops long, and 32 bits per as. 4TB of memory should be enough.
In reality you can drop it to /24, nobody advertises less than that. On average 16 million entries and say 30 hops, 2GB should be enough to store the routing table.
(Score: 3, Informative) by cafebabe on Saturday August 16 2014, @09:30PM
Here [soylentnews.org].
1702845791×2
(Score: 2, Insightful) by Anonymous Coward on Saturday August 16 2014, @09:33PM
Old gear has small routing tables. Cisco is using the recent internet-burp to scare people into upgrading.
Yawn...
(Score: 4, Insightful) by Ryuugami on Saturday August 16 2014, @10:09PM
Doubt it. With the recent NSA fallout, at this time people are more likely to upgrade away from Cisco.
If a shit storm's on the horizon, it's good to know far enough ahead you can at least bring along an umbrella. - D.Weber
(Score: 5, Interesting) by zocalo on Saturday August 16 2014, @10:15PM
Can't speak for other vendors, but at least in Cisco's case the bulk of their hardware can be fixed with a configuration change and a reboot. What I found particularly ironic about this though is that several of the ISPs that had problems are amongst those that create the problem in the first place by failing to aggregate their IP allocations in the BGP table properly. What goes around, comes around...
UNIX? They're not even circumcised! Savages!
(Score: 0) by Anonymous Coward on Saturday August 16 2014, @09:51PM
net neutrality requires to "route early and route often".
(Score: 2) by kaszz on Sunday August 17 2014, @01:44AM
Lets solve two problems at once with a upgrade. Get bigger routing table and get rid of NSA "features".
As a stop gap solution wouldn't a memory add-on be possible? like abusing the memory socket or other connection like when 386 or 486 got the offer to use overdrive CPUs that just exploited the socket.
Btw, if it's 512k entries, how many kB is this blob really? and is packets routed to interfaces or IPs?
(Score: 0) by Anonymous Coward on Sunday August 17 2014, @04:57AM
Remember that Cisco did not cooperate. The USA
did things the hard way. You think they can't do
that with any other vendor? You think other
governments won't do that? Suppose you are in
Germany and you buy Huawei gear from China. First
you get China's backdoors. Then the gear gets
shipped to France, where France adds one. Then
the gear goes to Germany, where you get another.
Finally, as a shipping company delivers it,
both the USA and Russia add backdoors. Enjoy!
(Score: 1) by maxwell demon on Sunday August 17 2014, @05:57AM
I'd say that's a design problem of the internet protocol. Previously I was not aware that every little ISP router has to store all the possible routes. That's a scalability issue. If they now increase the number of routes they can store, it will only be a temporary solution because the number of routes will grow, and more importantly, it will grow much faster than the number of routers.
I don't think it should be necessary for every router to store every route. All it should need to know is to which adjacent router to send which packet. How that router decides to further send the package should not be its issue.
The Tao of math: The numbers you can count are not the real numbers.
(Score: 2) by isostatic on Sunday August 17 2014, @06:39AM
Oh, why didn't we have you available earlier! You could have saved us all this problem!
One minor question, I'm sure you have an answer. I connect to 4 other networks, one in New York, one in Frankfurt, one in Singapore, one in London. I have a packet for 12.54.124.65. Which do I send it to?
(Score: 1) by dougisfunny on Sunday August 17 2014, @07:43AM
Come now man, that's an absurd way to set up a token ring.
(Score: 1, Informative) by Anonymous Coward on Sunday August 17 2014, @08:17AM
I am afraid you do not know how routing or IP works. This does not have anything to do with IP save for the numbering conventions. Routes can be aggregated. Say I have networks 17.0.0.0 /16, and 17.1.0.0 /16, they can be aggregated into a 17.0.0.0 /15 route. There are also default routes that everything, everything uses. That is, if there is no specific route then the traffic is dumped to the default route. Your computer has one, your wifi router has one, everything that is either default or set up correctly has one. You want to have the entire internet in one route, you can do it. Your local router is doing it right now with a 0.0.0.0 route to your ISP.
This failing is having routing tables that are way, way too large. That happens as an intentional failure to use the tools available.
(Score: 0) by Anonymous Coward on Monday August 18 2014, @03:37PM
I think the post you are responding to is basically saying the same thing as you.
(Score: 2) by maxwell demon on Tuesday August 19 2014, @06:07AM
OK, then it's not a case of bad design, but a case of bad naming. In normal language, a "route" is a complete end-to-end description of a way, so when I read that there are more routes than the routers have space to store, that's what I thought they store. I guess you agree that this would have been a design problem.
As you describe it, it's basically what I wrote how it should be. Now if those entries had been given a better name, say "direction", that misunderstanding would not have occurred.
The Tao of math: The numbers you can count are not the real numbers.
(Score: 0) by Anonymous Coward on Sunday August 17 2014, @09:13AM
Don't have the link because I read it earlier (on another device and well after the dupe article here). but why is this article quoting 512,000 more than once, I expected SN subs or editors of all places to be able to count to 512K a little more accurately.
(Score: 3) by tibman on Sunday August 17 2014, @05:09PM
512K looks like 512,000 to me. Maybe you thinking 512KB?
SN won't survive on lurkers alone. Write comments.
(Score: 0) by Anonymous Coward on Monday August 18 2014, @03:03AM
As I understand it, they're talking 512k (524,000-ish), not some weird base-ten number. If you want to do things fast on a computer, why would you use anything other than a power of two for your table size?