Web Hosting

1
Web Hosting

• Herng-Yow Chen

2
Outline

• How different web site can be virtually hosted
  on the same server, and how this affects HTTP
• How to make web sites more reliable under heavy
  traffic
• How to make web site load faster

3
Hosting services
4
A Simple Example Dedicated Hosting
Irenes ISP
www.joes-hardware.com
Content
Joe
client
www.cajun-gifts.com
Internet
www.marys-antiques.com
Content
client
Mary
www.irenes-isp.com
5
Virtual Hosting
6
Virtual Hosting
Content
Irenes ISP
www.joes-hardware.com
Joe
www.marys-antiques.com
Content
Mary
client
www.cajun-gifts.com
Internet
client
www.irenes-isp.com
7
Virtual server request lacks Host information
8
Virtual Server Request Lacks Host Information
(A getting http//www.joes-hardware.com/index.html
)
www.voting-info.gov www.joes-hardware.com www.mary
s-antiques.com
GET /index.html HTTP/1.0 User-agent SuperBrowser
v1.3
Internet
Client A
/voting /mary /joe
GET /index.html HTTP/1.0 User-agent WebSurfer
2000
Client B
HTTP/1.0 requests do not contain hostname
information, so they do not support web servers
that host multiple web sites. (HTTP/1.1 supports
a Host header to fix this problem)
(B getting http//www.marys-antiques.com/index.htm
l)
9
Making virtual hosting work

• Virtual hosting by URL path
• Virtual hosting by port number
• Virtual hosting by IP address
• Virtual hosting by Host header

10
Virtual hosting by URL path

• http//www.joes-hardware.com/joe/index.html
• http//www.marys-antiques.com/mary/index.html
• The problem?

11
Virtual hosting by port number

• Install multiple HTTP servers
• one listens 80
• one listens 81, and so on.
• The problem?

12
Virtual hosting by IP address

• Multiple IPs are assigned to single computer.
• Different IP corresponds to different host.
• The problem?

13
Virtual hosting by IP address
www.voting-info.gov209.172.34.2 www.joes-hardware
.com209.172.34.3 www.marys-antiques.com209.172.3
4.4
Dest IP address Directory 209.172.34.2
/voting 209.172.34.3
/joe 209.172.34.4 /mary
209.172.34.3
Client A
Internet
/voting /mary /joe
209.172.34.4
Client B
14
Virtual hosting by Host header

• HTTP/1.1 solution
• Use single IP, single port to serve incoming HTTP
  requests to different hosts.

15
Virtual hosting by Host header
(A getting http//www.joes-hardware.com/index.html
)
www.voting-info.gov www.joes-hardware.com www.mary
s-antiques.com
GET /index.html HTTP/1.1 User-agent SuperBrowser
v1.3 Host www.joes-hardware.com
Internet
Client A
/voting /mary /joe
GET /index.html HTTP/1.1 User-agent WebSurfer
2000 Host marys-antiques.com
Client B
The HTTP Host header carries the hostname
information that would otherwise be lost in
normal server requests,allowing name-based
virtual hosting
(B getting http//www.marys-antiques.com/index.htm
l)
16
HTTP/1.1 Host Headers

• Syntax and usage
• Missing Host headers
• Interpreting Host headers
• Host headers and proxies

17
Making web sites reliable

• There are several times during which web sites
  commonly break
• Server downtime
• Traffic spikes
• Network outages or losses
• Some ways of anticipating and dealing with the
  problems
• Mirrored server farms
• Content distribution networks

18
Mirrored server farms

• HTTP redirection
• DNS redirection
• Discussed in Chapter 20.

19
Mirrored Server Farms
Client
Client
Replica origin server
Client
Internet
Switch
Master origin server
Client
Client
Replica origin servers
20
Mirrored Server Farms (cont.)
New York Replica origin server
Chicago (HQ) Master origin server
Internet
Miami Replica origin server
Little Rock Replica origin server
21
Content distribution networks

• Surrogate caches in CDNs
• Proxy caches in CDNs

22
Proxy Caches in CDNs
Caching Proxy
Client
Internet and lots of origin servers
Client
Switch
Client
Client
23
Making web sites faster

• Many of the technologies mentioned in the
  previous slides also help web site load faster.
• Server farms and distributed proxy caches or
  surrogate servers distribute network traffic,
  avoiding congestion.
• Distributing the content brings it closer to end
  users, so that the travel time from server to
  client is lower. The key to speed of resource
  access is how requests and responses are directed
  from client to server and back across the
  Internet. (see more details in next lectures,
  redirection methods.)
• Another approach to speeding up web sites is
  encoding the content for fast transporting.
  Namely, compressing the content. (see Transfer
  Encoding and Chunked Encoding for details.)

24
Reference

• http//www.ietf.org/rfc/rfc3040.txt
• RFC 3040, Internet Web Replication and
  Caching Taxonomy, is a
• reference for the vocabulary of web
  replication and caching
• applications.
• http//search.ietf.org/internet-drafts/draft-ietf-
  cdi-request-routing-reqs-00.txt
• Request-Routing Requirements for Content
  Internetworking.
• Apache The Definitive Guide
• Ben Laurie and Peter Lauire, OReilly
  Associates, Inc.
• This book describes how to run the open
  source Apache web server.