wikipedia.org and wikimedia.org: how are they hosted?

3

0

Pinging either of them echoes back from an ip address of 198.35.26.96. So whatever PC's these two websites are hosted on, they are connected to the internet by a common router, which is then port-forwarded to either of these PC's, right? And both of these two forwarded ports are registered with the corresponding domain names wikipedia.org and wikimedia.org, right?

But ip-address.org shows that the ip 192.35.26.96 has Wikimedia Foundation as internet service provider, and text-lb.ulsfo.wikimedia.org as ip host. I didn't know Wikimedia Foundation would count as an ISP; I thought the ISP would be some other networking firm with a wireless tower here and there(you got the idea), who connects wikipedia with the world with a strong broadband connection. Also the IP host is a subdomain of wikimedia.org, which is equally puzzling. How could wikimedia.org be defined and registered even before the IP host had been defined and registered?

If my picture of the networking world is naive, please clarify why.

Tony Marshle

Posted 2016-12-23T10:42:29.640

Reputation: 49

Question was closed 2016-12-23T13:31:49.493

Answers

4

I think your question is a bit too broad. I will try to explain a bit in very general terms, so you hopefully get the gist. But I think reading more about networking in general would be beneficial as well if you're interested in that kind of stuff :)

how is wikimedia/wikipedia hosted?

Wikimedia's infrastructure and everything is on their wikitech page. Can get started on https://wikitech.wikimedia.org/wiki/Portal:Wikitech

So whatever PC's these two websites are hosted on, they are connected to the internet by a common router, which is then port-forwarded to either of these PC's, right?

In a general sense, there is typically no limit to the number of sites per computer (other than obvious processing limits). For Wikimedia, they use clusters of servers with caching, load balancing, etc. Even if the website resolves to one public IP, there can still be many different services provided behind what you connect to.

I didn't know Wikimedia Foundation would count as an ISP; I thought the ISP would be some other networking firm with a wireless tower here and there(you got the idea), who connects wikipedia with the world with a strong broadband connection.

Yep, Wikimedia Foundation is technically an ISP. They run an autonomous system with number of AS14907 (US) and AS43821 (EU), and their IP ranges are registered to their ASNs. An ISP does not need to be an extremely large company, nor does it need to provide any kind of backbone connectivity. In fact, many (most?) ISPs do not provide much of that connectivity. Wikimedia sits at their datacenters who provide their Internet connectivity and connectivity to multiple Internet Exchange Points. This lets you, the end user, get connected to Wikimedia.

Also the IP host is a subdomain of wikimedia.org, which is equally puzzling. How could wikimedia.org be defined and registered even before the IP host had been defined and registered?

I do not fully understand what you mean. You mean the reverse-DNS of 198.35.26.96 is "text-lb.ulsfo.wikimedia.org"? Reverse DNS records can be delegated to anything. I could set the reverse DNS of an IP I own to be "superuser.com", "not.a.valid.website.superuser", or basically anything.

Hopefully that cleared up a little bit of confusion, and some resources/ideas to proceed with :)

Iskar

Posted 2016-12-23T10:42:29.640

Reputation: 266

1

Some additions to the other answer:

Pinging either of them echoes back from an ip address of 198.35.26.96. So whatever PC's these two websites are hosted on, they are connected to the internet by a common router, which is then port-forwarded to either of these PC's, right?

In general, what you describe is possible, but most websites have grown a bit beyond that – very little about home/smb internet providers applies to large-scale hosting. (Compare e.g. a mp3 player and a professional recording studio.)

In the specific case of Wikimedia it's not just one PC per site anymore – it's several racks of servers storing the articles, the images, the wiki code, etc. (See their own webpage about that, or photos and blog posts on Google.) Chances are that Wikimedia's network is bigger than some ISPs.

But even though they have many servers, it's usually not a one-site-per-server setup – often a single server hosts many websites, and one IP address might be load-balanced between many servers. (The "many websites on one server" part is usually called virtual hosting. All HTTP requests include a Host: header telling the server which site you tried to access. That makes it possible to host even hundreds or thousands of sites on the same server, same address, same port – the HTTP service just looks up the matching configuration.)

I didn't know Wikimedia Foundation would count as an ISP; I thought the ISP would be some other networking firm with a wireless tower here and there(you got the idea), who connects wikipedia with the world with a strong broadband connection.

Taken literally, no, Wikimedia is not an Internet service provider. But such websites use the term 'ISP' rather broadly – e.g. in this case they probably mean an 'autonomous system' owner, which might be the Internet equivalent of a legal entity – you need to have an AS number to be able to own IP addresses and announce them to the rest of the Internet using BGP. Since most autonomous systems are ISPs, some websites just use the terms interchangeably.

That doesn't mean they're the same thing though; anyone (even an individual) can obtain an AS number after some paperwork. For example, many organizations do this as part of setting up a redundant connection – they obtain their own ASN, buy a block of IP addresses, and announce it through both ISPs at once – in essence, becoming part of the Internet exactly like ISPs themselves do.

Also the IP host is a subdomain of wikimedia.org, which is equally puzzling. How could wikimedia.org be defined and registered even before the IP host had been defined and registered?

Domain registration is entirely separate from IP addresses – buying a domain merely reserves the name, and then you configure where it should point. (Most domain registrars have a control panel where you can edit the domain and subdomains.)

Similarly, reverse DNS (mapping an IP address back to some domain name) can be changed at any time by the IP's owner.

user1686

Posted 2016-12-23T10:42:29.640

Reputation: 283 655