Slashdot Mirror
Is The Internet Growing Too Fast?
SpunOne writes: "According to this article, the Internet is growing faster than today's routers can handle. After years of predictable growth, the size of the routing table and traffic in it exploded during the past six months, topping 104,000 entries in March, compared with 75,000 a year ago. Even more troubling is evidence that frequent updates to the routing table entries by network managers are causing instability in the Internet's backbone routing infrastructure."
I did propose to the IPv6 group that two things in the new IP space be done. First, that a bit toggle whether this is a roaming/DHCP IP address (an opt-out system, plus additional space to indicate if you were even on earth, or in orbit, etc.), and the second, including a rough latitude and longitude for the system with the IP. It was never necessary nor possible to get an exact location for every device. Besides, everything in a rack would have the same IP.
When a pitched it then, it was quickly shot down. However, it allowed for regional routing, which is where you just sent packets to the router responsible for a major lat/lon, which then forwarded packets to things in its area. Really simple.
It also allowed you to figure out where things were coming from (if the user opted in), which meant you could redirect them to a regional mirror. As a side-effect, folks like Yahoo could actually keep all the French out of websites, since the IP would actually have some sort of location information in it.
Routing would be more efficient, and smaller, and faster. But hey, I'm just a nutcase, as the IPv6 guys said.
the fact here is that a multihomed router taking full BGP routes from their upstreams today needs atleast 256 megs of RAM and a big cpu (example cisco 7500 MINUMUM) in order to handle it..
The problem is all the recent "hey let's be a DSL provider!!" idiots that have no understanding of route summarization and addressing.. Thankfully many of these ISPs are going bankrupt now so hopefully all their /23's and /24's will be withdrawn after their lights go out.
The ISPs need GOOD engineers who can design a network in a manner that only the most aggregate route will be announced to external peers.. the biggest problem is they assign smaller blocks and then that customer gets multihomed and the original provider has to cut holes in their BGP filters in order for it to work right.
If you're going to multihome, GET YOUR OWN ADDRESSES FROM ARIN!!! Places like Verio have very Nazi-ish filtering policies and their routing tables only hit about 85,000.. While verio makes bad routing decisions, this shows an example of what the internet routing table SHOULD look like.
SUMMARIZE, SUMMARIZE, FUCK YOUR TOASTER, IT DOESN'T NEED AN IP ADDRESS!@@!#@!#
You're absolutely right that our ability to cope with the net is what's suffering. A friend of mine once said that the Internet is "like giving +1 to everyone's intelligence. If they don't know something, they now have a tool that lets them look it up and get a ton of information quickly." But now we realize that you only get the +1 bonus if you are already fairly intelligent, because it takes intelligence to be able to do Internet research, more intelligence to determine if the sites that you're looking at are good sites and/or responsible with facts.
With the wisdom of hundreds of other netters, and the net's gift of great communication, I can be more intelligent than a doctor is about a given prescription drug, for example. But if I don't select sources correctly, I can be downright dangerously ignorant about that drug.
If the tools that we have for research improve, THEN we can add +1 to everyone's intelligence. So what we have in the meantime is a tool where the smart get smarter and the dumb stay pretty much where they are.
It may be that the current economic contraction may be just what the net needs. It may stagnate in growth for a little bit -- and then we'll have a little bit of time to catch our collective breaths and allow humans to catch up with what is possible. One reason we have all these fucked companies is that the masses didn't adapt as quickly as the netted elite. The main reason is the temporarily insane marketplace, sure, but a lot of online things are much better than their offline counterparts, it's just that people couldn't catch up with the growth of the net. It's better to meet online than to travel 2000 miles to meet, but today we are still adapting to the idea that we can shop for airline tickets over the net.
The net has allowed us to see what's possible and to implement new and better approaches. But it could not hold our hand and show us the new and better approaches. With time at a premium in everyone's lives, no-one can afford the time for experimentation or learning with these new approaches.
OK, I'll stop now before I become Katz.
This is the single funniest statement I've read in days.
They blame multihoming - however there is a limit of 65535 possible autonomous systems out there under BGP4, so if each of these has an average of 3 entries, doesn't that give a max number of route entries of under 200,000? Except you also have to multiply by the number of IP address ranges (networks and subnets) advertised by each AS - but still there would seem to be a limit not too far from where we are now. Which probably does mean it's time to replace BGP...
Energy: time to change the picture.
This was posted on NANOG this morning and should be required reading. It is from Sean Doran who basically built Sprint's Network in the early/mid 90s and is probably *the* authority on this kind of stuff. Read it.
I see everyone blaming the small multihomed sites and talk about banning multihoming on this forum, I strongly disagree with that and I strongly disatree this nasty quote from the article:
"Half of the companies that are multihomed should have gotten better service from their providers," says Patrik Faltstrom, a Cisco engineer and co-chair of the IETF's Applications Area. "ISPs haven't done a good enough job explaining to their customers that they don't need to multihome."
Yeah sure, my provider told me how I didn't need to multi-home and I got burned. Excuse me for stamping on your elitism here, but everybody wants redundancy and you shouldn't have to be a fortune 500 to get it.
About a year and a half ago our company was looking into upgrading bandwidth, and since we already had a t1 my boss figured we could buy another t1 one from a different company than our normal bandwidth provider, thus achieving increased speed redundancy with a different isp at the same time.
Well I found out that in order to do something like that I needed to run BGP to get myself into the core routers routing table because I would have multiple paths to my network. I was a big concerned about this because it sounded drastic to me and I spoke with my tech rep at my current provider and we mutually agreed that perhaps it wasn't the best thing to do as we could get redundnacy from their network and I wouldn't need to have my own AS.
Well, to make a long story not quite as long, I got burned. They had some routing problems that affected both of my links at once, and we were down for a bit. I had to explain to the boss why I had made the decision I did and he wasn't real happy about it. I am about to be installing our 3rd t1 and switching to bgp so we can be multihomed.
This is a very typical story, and is the primary reason that the BGP4 tables are so huge. Every dotcom and their mom has been going through a similar scenario the past year or so. Now of course this is starting to not work very well so we are seeing some problems, and these same dotcoms are being blamed for this.
The problem isn't the dotcom's, its due to limitations in the current system. I suppose they (we) do shoulder some of the blame, but christ shouldn't we be allowed to have some kind redundancy? What is there some kind of special "VIP's only" sign in front of the redundancy bar? To hell with that. Obvously the current situation cannot continue, but I'll tell you right now that all of these dotcom's (and their moms) are not going to be giving up redundancy, so you core router guys better figure out why to let everyone in on the redundancy bandwagon.
Sigs are awesome huh?
Actually, I'm pretty sure that Cisco's 2500-series routers to have m68k processors. 68030, I believe. Everything never has a PowerPC variant. Dunno about Juniper, but I'm going to guess they're running a RISC processor of some sort.
The hardware issue was another factor covered by NANOG this morning in response to this article. The long and the short of is that throwing hardware at a problem might get you through the day, but good planning and forethought will bring you through a lot more cleanly.
With any router manufacturer, your value-added is never the hardware but rather the software (i.e. Cisco's IOS) and the service. Sure, you could grab a PII and run GNU Zebra for your BGP peering, but the CCO is a great resource to have when troubleshooting. Puts MS's TechNotes system to shame. It WORKS. Plus, working in downtown Chicago I know that if my 7200 takes a dive, Cisco's got a parts depot blocks away where I can get parts.
-carl. We've got computers, we're tapping phone lines, you know that ain't allowed - Talking Heads, "Life During Wartime"
Second of all, a greater concern might be "Is the Internet growing too disorganized?" There are ten jillion pages out there, and the vast majority of them aren't even linked to from other documents. They don't show up on search engines, they just sit there, with the web masters wondering why they've only gotten 3 visits in the past year.
Even the sites that can be found by search engines are getting increasingly hard to organize. Yahoo! is starting to wobble in their traditional high-quality, hand-picked links directory. They can't keep up with the net, so they've started implementing pay-for-listing programs. The Open Directory Project survives because of heaps and heaps of volunteer editors, but every category varies in qualities, and some basic categories don't even have editors. Many other search engines have attempted dynamically-created directories based on keywords, but these are easy to spam and often have very low-quality content.
All the disorganization also affects our information processing skills. We don't read like we used to. Hardcore web surfers are generally incapable of sitting down and enjoying a good book, because they're too accustomed to the "This page doesn't have it, go to the next one" cognitive paradigm.
What we really need is a new way to organize web sites, perhaps based on a combination of client (most visited sites), server (author-specified categories), and parsed (most linked-to sites) information.
The internet is not growing too fast. Our ability to cope with it, however, is failing to grow with it.
Got Rhinos?
Ignore this shit. Every year, someone predicts the internet will "collapse under its own weight". Guess what? It NEVER DOES. People have been claiming the sky is falling since NSFnet became available to the public -- I'm still waiting.
I think you misspelled "Feel free to sit back on your lazy ass while somebody else designs and implements a fix - just like they always do". HTH. HAND.
Any sufficiently well-organized community is indistinguishable from Government.
There is also some very interesting information regarding IPv6 in various sites, such as 6BONE's, and Sun's. It is really great to poke around with IPv6 stuff, there are a lot of programs that support it by now, such as lynx (-dev tree only), w3m, BitchX, epic, etc. etc. etc. And also, IPv6 is cool because it lets you create such educational hosts like dead:beef:c0ff:eeca:bf00:3:133:7.
If you don't believe me, here is my sit1 interface:
You're tired of Slashdot ads? Get junkbuster now!
Yeap, seen this too many times to be funny: Network managers, when you make a change to the networks you advertise or your filters, use "clear ip bgp * soft out"! Without the soft out, all your routes get withdrawn and then re-added around a minute later, crating a mini route wave through the Internet. Instead of forcefully resetting your peers, soft out will just add to the table version without withdrawing routes. You could also use "soft in" for applying filters to incoming routes also BUT watch out for memory useage with this command. Soft out doesn't have this issue.
This is fairly basic Cisco IOS stuff but I've seen network admins from 2 of the top ISPs do this on peering point routers that were advertising a lot of routes to a lot of peers.
A journey of a thousand miles starts with a brutal anal raping at airport security
To avoid a complete meltdown of the Internet, President G.W. Bush has recommended that rolling blankouts begin within the next two months. During peak periods of Internet traffic, up to 100,000 sites may have their routing tables blanked for up to an hour in order to reduce the load on the system.
"I may not know much about the Internet, but I don't think you need a PhD to see what all this free music trading is doing to our nation supply of routing tables," Bush was quoted as saying. Asked if he was implying that Napster was responsible for the rolling blankouts, he replied "The RIAA says they are, so it must be true."
According to this post it's not growing fast enough...
Your pizza just the way you ought to have it.