Internet Statistics

1
Internet Statistics Facts and Figures

• 1st Tutorial Session for CEG3180B
• January 18th, 2005

2
First things first Internet?

• Probably the most used example when it comes to
  Wide Area Networks (WANs)
• Originally a DoD project 1, it has become the
  most widely used public internetwork
• It is the most technologically heterogeneous
  network ever to exist, short for one aspect
  TCP/IP

3
Why Internet Statistics?

• Question Why would we be interested in the
  Internet host count trends?
• Several pertinent answers
• Curiosity ?
• Bandwidth consumption (i.e., traffic) forecasting
• Address space usage forecasting
• etc.

4
Well, how many are they?

• How many what?
• Hosts
• Originally, a host was a single computer on the
  network
• With the introduction of virtual hosts, however,
  a single computer could represent more than one
  host
• So should we count virtual hosts too or not?
• According to the ISC, in June 2004 there were
  about 300 million hosts in the Internet

5
(No Transcript)
6
What does this survey tell us?

• Basically, that the number of hosts in the
  Internet is growing exponentially
• This means that the Internet population doubles
  approximately
• every 15 months!

7
How does the survey work?

• It is a Domain Name System (DNS) 3 survey
• What is the Domain Name System?
• A hierarchical (i.e., similar to the postal
  address system) way to name hosts in order to set
  a correspondence between domain names and IP 4
  addresses
• Top-level domains .com, .net, .edu, .mil, .gov,
  .org, .int (and the more recent .biz, .info, .to
  etc.) and the country top-level domains
• Second-level domains usually (but not always)
  organizations
• E.g., www.uottawa.ca means the web server (www)
  of the University of Ottawa (uottawa) which is
  in Canada (ca)

8
How does the survey work? (2)

• Two methods
• Walking the domain name tree and doing zone
  transfers of domain data in order to discover
  hosts and further subdomains (original method,
  used until 1997)
• Walking the reverse delegation zones (i.e.,
  in-addr.arpa) and counting all addresses that
  have been assigned a domain name (new method,
  used since 1998)

9
Is the survey accurate?

• Question How far can we trust these results?
• Answer They will have to do it.
• Neither of the two survey methods is free of
  errors it is virtually impossible to know the
  exact number of hosts in the internet
• However, a good approximation should be sufficient

10
Is the survey accurate? (2)

• Reasons for the surveys not being able to provide
  error-free results
• Many DNS servers do not allow for whole zone
  transfers
• Not all hosts are registered in a domain name
  server
• Poorly configured DNS servers can lead to bogus
  entries
• Just because a hostname is assigned an IP address
  or vice-versa, does not mean that the host
  actually exists
• Poor connectivity can pose difficulties to the
  data collection process

11
Interesting facts

• What has the number of hosts in a certain country
  to do with its economic status?
• More .net hosts than .com hosts does this mean
  that there are more hosts for infrastructure
  than for services?
• The top 2 host names 1 million www, 400k
  mail
• Canada has about 3.5 million hosts in 21k domains

12
Conclusions

• The Internet population is growing at an
  exponentially rate
• Keeping track of the number of hosts in the
  Internet is important, for several reasons
• However, an extremely accurate counting is
  virtually impossible

13
References

1. http//www.isoc.org/internet/history
2. http//www.isc.org/ds
3. P. V. Mockapetris, STD0013/RFC1034 Domain names
   concepts and facilities
4. J. Postel, STD0005/RFC0791 Internet Protocol
5. S. Deering, R. Hinden, RFC2460 Internet
   Protocol, Version 6 (IPv6) Specification