1. Web Servers

2. Pre-lecture Survey: What is the #1 web server:
Apache
Google
MS IIS HTTP server
nginx
Sun
Other

3.
Generic Overview

4. Web Servers A web server can be a: Computer Program Computer
Responsible for accepting HTTP requests from clients (web browsers)
Returns HTTP responses with optional data contents
Usually web pages
HTML documents
Linked objects (images, etc.).
Computer
Running a computer program which provides the above functionality

5. Common Features

6. Common Features HTTP Accepts HTTP requests from a client
Provides HTTP responses to the client
Typically an “HTML” document can be:
File containing HTML statements
Raw text file
Image
Some other type of document
defined by MIME-types
If an error in a client request or trying to service the request:
Web server sends an error response
May include custom HTML
May have text messages
Better explain the problem to end user

7. Common Features Logging
Web servers keep detailed information to log files
Client requests
Server responses
Allows the Webmaster to collect data
Running log analyzers

8. Additional Features Authentication Handles:
Optional authorization before allowing access to some or all resources
Requires a user name and password
Handles:
Static content
Dynamic content
Support one or more related interfaces
SSI, CGI, SCGI, FastCGI, JSP, PHP, ASP, ASP .NET, Server API such as NSAPI, ISAPI, etc.

9. Additional Features HTTPS support Content compression VIA SSL or TLS
Allows secure (encrypted) connections
Uses port 443 instead of port 80
Content compression
I.e. by gzip encoding
Reduces the size of the responses
Lower bandwidth usage, etc.

10. Additional Features Virtual hosting Large file support
Serve many web sites using one IP address
Large file support
Serve files greater than 2 GB
Typical 32 bit OS restriction
Bandwidth throttling
Limit the speed of responses
Do not saturate the network
Able to serve more clients

11. Origin of the returned content
Where does the requested material come from?
Origin of the returned content

12. Content Origin Origin of the returned content may be:
Static
Pre-existing data file
Contents loaded on request
Dynamic
Content generated by another program
Script (programming language)
Creates/retrieves the requested information
Static content is usually delivered much faster than dynamic content
2 to 100 times
Especially if the latter involves data pulled from a database

13. How does it find it?
Path translation

14. Path translation
Web servers map the path component of a Uniform Resource Locator (URL) into:
Local file system resource
Static requests
Internal or external program name
Dynamic requests
For a static request the URL path specified by the client is relative to the Web server's root directory
This is not the same as the computers root directory

15. Path translation
Consider the following URL requested by a client Web Browser:
Client's Web browser translates it:
Where
Use the HTTP protocol
The Web server to connect to
This is translated to an IP address by DNS
Sent to 93: :80
/path/file.html
The resource to access
Generates the following HTTP 1.1 request sent to the IP address:
GET /path/file.html HTTP/1.1 Host:

16. Path translation (cont.)
Web server host (
Sees the request is for port 80
Sends request to the Web Server software
Appends the given path to the path of the servers Web root directory
On Unix machines typically /var/www/htdocs or /var/www
Result would then be the local file system resource:
/var/www/htdocs/path/file.html
Web server:
Retrieves the file, if it exists
Processes it by the Web servers rules
Sends a response to the client's web browser
Response:
Describes the content of the file
Contains the file requested or a response

17. Performance

18. Performance Web servers must:
Serve requests quickly!
From more than one TCP/IP connection at a time
Some main key performance parameters are:
number of requests per second
depends on the type of request, etc.
latency response time in milliseconds
for each new connection or request
throughput in bytes per second
Depends on
File size
Content cached or not
Available network bandwidth
etc.
Measured under:
Varying load of clients
Varying requests per client

19. Performance
Performance parameters may vary noticeably depending on the number of active connections
Concurrency level
Fourth parameter supported by a web server under a specific configuration
Specific server model used to implement a web server program can bias the performance and scalability level that can be reached under heavy load or when using high end hardware
many CPUs, disks, etc.

20. Load limits

21. Load limits Web servers have load limits
Can be set in a configuration file
Can handle only a limited number of concurrent client connections per IP address (and IP ports)
Usually between 2 and 60,000
Default between 500 and 1,000
Can serve only a certain maximum number of requests per second depending on:
Settings
HTTP request type
Content origin
Static
Dynamic
Served content cached or not
Hardware and software limits of the native OS
A web server near or over its limits
Becomes overloaded
Unresponsive

22. Overload causes

23. Overload causes
A sample daily graph of a web server's load, indicating a spike in the load early in the day.

24. Overload causes At any time web servers can be overloaded because of:
Too much legitimate web traffic
Thousands or even millions of clients hitting the web site in a short interval of time
DDoS
Distributed Denial of Service attacks
Coordinated
Computer worms
Abnormal traffic because of millions of infected computers
Not coordinated
XSS viruses
Millions of infected browsers and/or web servers
Internet web robots
Traffic not filtered / limited on large web sites with very few resources (bandwidth, etc.)
Internet (network) slowdowns
Client requests are served more slowly and the number of connections increases so much that server limits are reached
Web servers (computers) partial unavailability
Required / urgent maintenance or upgrade
HW or SW failures
Back-end (i.e. DB) failures, etc.
Remaining web servers get too much traffic and they become overloaded

25. Overload symptoms

26. Overload symptoms Symptoms of an overloaded web server include:
Requests are served with (possibly long) delays
from 1 second to a few hundred seconds
500, 502, 503, 504 HTTP errors returned to clients
Sometimes also unrelated 404 error or even 408 error may be returned
TCP connections are refused or reset (interrupted) before any content is sent to clients
In very rare cases, only partial contents are sent
This behavior may well be considered a bug
Even if it stems from unavailable system resources

27. Anti-overload techniques

28. Anti-overload techniques
To partially overcome load limits and to prevent overload use techniques like:
Managing network traffic by using:
Firewalls
Block unwanted traffic
Bad IP sources
Bad patterns
HTTP traffic managers
Drop, redirect or rewrite requests having bad HTTP patterns
Bandwidth management and traffic shaping
Smooth down peaks in network usage
Deploying web cache techniques
Use different domains to serve different content (static and dynamic) by separate Web servers, i.e.:
Serves static images
Serves dynamic data requests

29. Anti-overload techniques
Techniques continued:
Use different domain names and/or computers to separate big files from small/medium files
Be able to fully cache small and medium sized files
Efficiently serve big or huge (over MB) files by using different settings
Using many Web servers (programs) per computer
Each bound to its own network card and IP address
Use many Web servers that are grouped together
Act or are seen as one big Web server
See Load balancer

30. Anti-overload techniques
Techniques continued:
Add more hardware resources
RAM, disks, NICs, etc.
Tune OS parameters
Hardware capabilities
Usage
Use more efficient computer programs for web servers, etc.
nginx
Use workarounds
Specially if dynamic content is involved

31. Historical notes

32. Historical notes World's first web server
Tim Berners-Lee proposed to CERN a new project
Ease the exchange of information between scientists
Using a hypertext system
Berners-Lee wrote two programs:
Browser
WorldWideWeb
Web server
Ran on NeXTSTEP

33. Historical notes First web server in USA Between 1991 and 1994:
Installed December 12, 1991
Bebo White at SLAC
After returning from a sabbatical at CERN
Between 1991 and 1994:
Simplicity and effectiveness of early technologies used to surf and exchange data through the World Wide Web helped to:
Port them to many different operating systems
Spread their use among lots of different social groups of people
First in scientific organizations
Then in universities
Finally in industry

34. Historical notes
1994: Tim Berners-Lee constituted the World Wide Web Consortium (W3C)
Regulate the further development of the many technologies in a standardization process:
HTTP
HTML
etc.
Following years saw an exponential growth of the number of web sites and servers

35. Software

36. Software
There are thousands of different web server programs available
Many specialized for very specific purposes
About 50 mainstream
The fact that a web server is not very popular does not necessarily mean
Lot of bugs
Poor performance
See Category:Web server software for a longer list of HTTP server programs.

37. Statistics

38. Statistics
Most popular web servers, used for public web sites, are tracked by
Netcraft.com
Details given by
Netcraft Web Server Reports
According to this site:
Apache has been the most popular web server on the Internet since April of 1996
July 2010 Netcraft Web Server Survey:
54.90% web sites on the Internet use Apache
25.87% web sites use IIS

39. Web Servers Developer September 2010 Percent October 2010 Change
Apache
129,782,948
57.12%
135,209,162
58.07%
0.95
Microsoft
54,787,167
24.11%
53,525,841
22.99%
-1.12
Google
15,312,751
6.74%
14,971,028
6.43%
-0.31
nginx
12,779,550
5.62%
14,130,907
6.07%
0.44
lighttpd
1,818,032
0.80%
1,380,160
0.59%
-0.21

40. Post-survey: What is the #1 web server:
Apache
Google
MS IIS HTTP server
nginx
Sun

41. Summary

42. Summary Concentrated on HTTP servers
Apache and IIS are the main web serving tools
nginx is rising fast
Apache still king
Currently declining
IIS Up and down, wandering
Usage tracked
Netcraft Web Server Survey