Page 1 of 1
How internet works
Posted: Thu Jul 27, 2006 1:06 am
by matthijs
Ok, sounds like a silly newby question, isn't it? Well, it's not. Ok, maybe it is.
Just know I have a lot of trouble connecting to some websites. Not all, but many. maybe 1/3 or 1/2 of sites I try to access don't respond (connection has timed out). I tested my own connection and all is fine (upload 4412 kbps, download 570 kbps, 848.0 Conn/min). Devnetwork flies as always. Sitepoint is not accessible.
So how can this be? It isn't the connecntion from my house to the cable. Is internet traffic clustered, and therefore if one cluster is having problems a whole set of websites is not accessible? Could it be a problem in my region?
Re: How internet works
Posted: Thu Jul 27, 2006 2:19 am
by Chris Corbyn
matthijs wrote:Ok, sounds like a silly newby question, isn't it? Well, it's not. Ok, maybe it is.
Just know I have a lot of trouble connecting to some websites. Not all, but many. maybe 1/3 or 1/2 of sites I try to access don't respond (connection has timed out). I tested my own connection and all is fine (upload 4412 kbps, download 570 kbps, 848.0 Conn/min). Devnetwork flies as always. Sitepoint is not accessible.
So how can this be? It isn't the connecntion from my house to the cable. Is internet traffic clustered, and therefore if one cluster is having problems a whole set of websites is not accessible? Could it be a problem in my region?
It's quite literally like a web. To get to any one server you need to go via many others. Try using the program tracert in windows or traceroute in linux. That will show you how many servers you go through to get here for example.
You send a request to DNS server first and foremost, that could possibly have issues itself and not know the IP of the server. If that responds with an IP, you then start looking for a route to the server, at which point you'll sort of hop from zone-to-zone until you end up there. The further away the server is, logically the more hops you'll make. Think of it like this: you don't have a direct cable between sitepoint and devnetworks.

Posted: Thu Jul 27, 2006 2:22 am
by daedalus__
if you do a search on google for something simple like 'how does the internet work' or look up some stuff on HTTP you can find some pretty simple or pretty complicated explantions. this one time, when i was younger and nubber i was looking for an answer to the question and i came across a 40-some page paper about http and how the internet works :O
Posted: Thu Jul 27, 2006 2:32 am
by SomeOne
are you one Win...system?? have you change MTU to 1400?

Posted: Thu Jul 27, 2006 3:44 am
by matthijs
Thanks for the input. I do basicly know how the internet works (read Shiflett's book on http), but I find it strange that today a lot of sites do not load, while others do. And it's not just a random occasional site which is offline. There are many.
So maybe I should refrase my question: is it possible that a whole cluster of websites is inaccessible because of some failure somewhere in a network centre?
@SomeOne: please explain what you mean. I am on a windows system, yes. But what is MTU?
Posted: Thu Jul 27, 2006 4:06 am
by Chris Corbyn
matthijs wrote:So maybe I should refrase my question: is it possible that a whole cluster of websites is inaccessible because of some failure somewhere in a network centre?
Yes. How can I explain this... Hmmm. OK, you have different routes to different locations.
For example, imagine you've set up intranet in your house, and each computer has access to the internet.
Now, to get the Jim's computer downstairs with IP address 192.168.1.4 from your own PC, where do you go?
2 hops: PC -> Router, Router -> Jim
Now, you ask for Kayleigh's PC with IP 192.168.1.6. Same thing, 2 hops. Now if your router was down you'd never make that first hop so you wouldn't reach kayleigh or jim, but kayleigh and jim's PCs may be functioning completely fine. It's the same situation with the internet but on a larger scale.
Let's say you ask for Google. How many hops? Let's see?
Code: Select all
cac @walnut ~$ traceroute www.google.com
traceroute: Warning: www.google.com has multiple addresses; using 64.233.183.104
traceroute to www.l.google.com (64.233.183.104), 30 hops max, 46 byte packets
1 nodns (192.168.2.2) 0.945 ms 0.903 ms 1.071 ms
2 dhcp-6-4 (192.168.6.4) 0.093 ms 0.115 ms 0.127 ms
3 border-router (213.205.138.153) 1.498 ms 1.411 ms 1.241 ms
4 s6-1-5-0.ar01.sbc.bb.pipex.net (212.241.252.46) 1.744 ms 1.790 ms 1.729 ms
5 g0-2-790.cr01.sbv.bb.pipex.net.132.72.62.in-addr.arpa (62.72.132.186) 1.998 ms 2.076 ms 2.104 ms
6 g0-2.cr04.tn5.bb.pipex.net (62.72.156.106) 7.632 ms 7.200 ms 7.230 ms
7 g3-25-6.cr05.tn5.bb.pipex.net (62.72.140.133) 7.500 ms 7.541 ms 7.852 ms
8 195.66.224.125 (195.66.224.125) 7.880 ms 7.841 ms 7.600 ms
9 216.239.43.91 (216.239.43.91) 16.745 ms 16.930 ms 25.718 ms
MPLS Label=176270 CoS=0 TTL=127 S=0
10 72.14.232.141 (72.14.232.141) 18.714 ms 19.129 ms 18.597 ms
11 72.14.233.77 (72.14.233.77) 19.225 ms 18.514 ms 18.460 ms
12 216.239.43.34 (216.239.43.34) 21.733 ms 27.263 ms 18.979 ms
13 64.233.183.104 (64.233.183.104) 19.485 ms 25.139 ms 18.838 ms
13 hops. If any one of those 13 servers is down I have no route to Google. Notice that at pipex alone I made at least 4 hops. I'm willing to bet your ISP are having a routing problem.
Posted: Thu Jul 27, 2006 4:30 am
by matthijs
Thanks for the explanation chris. Is quite logical.
And indeed it seems there has been a problem with a server/network point somewhere. Just 5 mins ago I tried reloading the sites I couldn't access (Sitepoint, news sites, etc) and suddenly they all loaded fine again.
I would think that if a certain hop is blocking the traffic, an alternative route should be available, isn't it? Because while 50% of the sites were accessible, the other 50% wasn't? But that means it has been my own ISP? A problem in a local hop?
Interesting stuff

Posted: Thu Jul 27, 2006 5:12 am
by Chris Corbyn
matthijs wrote:I would think that if a certain hop is blocking the traffic, an alternative route should be available, isn't it? Because while 50% of the sites were accessible, the other 50% wasn't? But that means it has been my own ISP? A problem in a local hop?
Interesting stuff

Yep, this is usually the case. You set up a cluster of servers and can use any one of them. The problem is, if the gateway to this cluster goes down (one server!) then it will fail. This gateway could be something referred to as an LVS which is basically a load balancer directing traffic to the individual servers in the cluster.
Another way clustering can be handled is that, at the name resolution end of the scale the name can resolve to multiple IP addresses. This is what's called "round-robin" and is what Google appear to do when you ask for google.com, and end up on one of their many masses of servers.
Code: Select all
traceroute: Warning: www.google.com has multiple addresses; using 64.233.183.104
Posted: Thu Jul 27, 2006 5:31 am
by SpiderMonkey
How the internet works?
Well, what you've got to understand is its not like a truck you can just dump stuff on. Its a series of tubes....
(Sorry, couldn't resist

)
Posted: Thu Jul 27, 2006 5:34 am
by Benjamin
SpiderMonkey wrote:How the internet works?
Well, what you've got to understand is its not like a truck you can just dump stuff on. Its a series of tubes....
(Sorry, couldn't resist

)
Oh yeah.. that is funny. My girlfriend and I watched that video.
Posted: Thu Jul 27, 2006 7:44 am
by matthijs
Yeah, I had to think about it as well when posting my question
For anyone who doesn't know what we are talking about:
http://www.youtube.com/watch?v=DClkE64nFDY
Posted: Thu Jul 27, 2006 10:20 am
by Jenk
If a link between continents goes down, you may well find half of the virtual world will disappear.
A few years ago one of the links between Europe and the US went down for a few hours. Certain ISP's customers literally could not access any US hosted sites.
Posted: Thu Jul 27, 2006 10:49 am
by Weirdan
A few years ago one of the links between Europe and the US went down for a few hours. Certain ISP's customers literally could not access any US hosted sites.
Yeah, I remember that, we had problems accessing our servers in US datacenters for like a 10 hours or so... man, that was scary

Posted: Thu Jul 27, 2006 11:54 am
by daedalus__
best video ever lmfao
Posted: Fri Jul 28, 2006 12:41 pm
by SomeOne