15.

561
Information Technology Essentials

Session 11
Under the hood of
a commercial website

Acknowledgments:

Adapted from Chris Dellarocas, U. Md..

Copyright © 2005 Thomas Malone, Chris Dellarocas

 Outline

• Issues for building enterprise-class systems

• Typical commercial website architecture: Travelocity

• How much does it cost?

 The story so far…

Web CGI/ASP
server script Database

Static
Internet
content
Webbrows
er
 Issues for building enterprise-level
eCommerce sites
• Security
– Physical security
– Access control
• Availability/Fault Tolerance
– Ensure the computer services remain available to users in the
face of partial failures
• Performance/Scalability
– Ensure that response time remains acceptable as site traffic
grows
• Content management
– Ensure that content development is done in a streamlined and
orderly fashion
 Enterprise-level eCommerce sites:
Summary of Technologies
• Security
– Physical security: Hosting
– Access control: Firewalls
• Availability/Fault Tolerance
– Replication
• Performance/Scalability
– Replication
– Load Balancing
– Web Caching
• Content management
– Content Management Software
Issue # 1: Network and Physical Security
 Types of firewalls

• Packet filter: Looks at each packet entering or leaving

the network and accepts or rejects it based on user-
defined rules.

• Application gateway: Applies security mechanisms to

specific applications, such as FTP and Telnet servers.

• Proxy server: Intercepts all messages entering and

leaving the network. The proxy server effectively
hides the true network addresses
 Web Hosting

• Ensure 24x7 site operation

• Provide access to network bandwidth

• Provide physical site security

 The various flavors of hosting

• Free hosting
– Your site on a shared server. Usually a shared domain name. Usually
supported by advertising

• Shared hosting
– Your site (and domain name) on a shared server in a special facility with
staff to respond when machines go down, etc. Often bundled with email.

• Dedicated hosting
– Your site(s) on a separate, dedicated server in a special facility. Varying
amounts of additional services (such as design of server configuration,
firewalls, replication, etc.)

• Collocated hosting
– Similar to dedicated hosting except you own the server rather than renting it.
 There is no single right answer

• Select level of hosting based on what kind of

company you are
– basic eCommerce presence ‚ shared hosting

» Century 21

– some in-house expertise, high volume ‚ dedicated hosting

» Land’sEnd, Vanguard
– eCommerce pioneer ‚ collocated hosting

» Yahoo, Amazon

 Issue # 2: Ensuring Availability and

Fault Tolerance

• Why do computers crash?

– Hardware errors
– Operating system errors
– Application errors
– Human errors

• Use redundancy to restore normal operation after crashes

– Data redundancy
– Active Replication
 Data Replication

• Keep several copies of same data (replicas)

• If one server is down, query next server

• Can improve response when load is heavy

• Problem: How to synchronize replicas?

 Active Replication

• Establish redundant copies of vital programs and

servers
– process groups
– every group member operates on its own replica

• Every message is processed by all group members

– members remain in mutually consistent states

• If one member fails, other members can still respond

 Issue # 3 Scalability:
Why you should care

• … the No.1 reason that customers got fed up and took

their business elsewhere was technical problems,
including unacceptably slow response times.
– Fortune magazine

• … 28% of Netizens that encountered glitches, left the

site never to return
– Business Week
 Technological Alternatives

• Local load balancer

• Site mirroring

• Content routing
 Load balancing solutions
Companies: Cisco, HydraWeb

Distributes client requests among

replicated servers Web
Serv
er
A
Load
balancer
Web
Serv
er
B

Web Lclod
oaLa
Internet Repcted
la
i
brows
er bl
aanc
er Webserv

ers
 Intelligent content routing

STEP 1

STEP 2

CUSTOMER
WEB SERVER

STEP 4 STEP 3
STEP 1 User enters standard URL

STEP 2 Web Server returns HTML with embedded URLs

pointing to Akamai Network

STEP 3 Rich content served locally

STEP 4 User browser request embedded objects

AKAMAI SERVER

Figure by MIT OCW. Companies: Akamai, Sandpiper

 Akamai network
• Over 15,000 servers in over 65 countries

AKAMAI SERVERS

Figure by MIT OCW.

 Issue # 4: Large-scale content
development
• Large number of authors contributing site content

• Diverse types of content (e.g., image, video, and other

media files)

• Need for regular content posting and replacement

(i.e., weekly sales promotions)

• Often one or more approvals are required before

content is posted

• Some content needs to be personalized or tailored to

match the needs and interests of a site visitor
 Additional issues

• Often support for multiple languages and time zones

is required

• Content presentation must consistently conform to

branding and appearance standards

• Version archiving and an audit trail

• Content must be viewable across a variety of

browsing devices, not just PCs
Solution: Content management systems
• Automatic support for content management workflows

 Enterprise-level eCommerce sites:

Summary of Technologies
• Security
– Physical security: Hosting
– Access control: Firewalls
• Availability/Fault Tolerance
– Replication
• Performance/Scalability
– Replication
– Load Balancing
– Web Caching
• Content management
– Content Management Software
 A Three Layer Website Architecture

BROWSER

INT
ERN
ET
Firewall

Load Balancer Web interface/

Front-end Manage page hits

Web Server Web Server

Session
Customer and session
Database Management
management
Middleware intelligence
software Transacion and requests
management

Read/
Back-end Inventory
Database
Content
DataBase
Profile
DataBase Store data
 How much does it cost?

• Typical e-Commerce corporate site

• 600,000 unique visitors per month

• Growth aimed at 3 million visitors per month in two

years
 Typical Hardware Costs
Startup Costs

Servers $129,000 33%

Network Hardware $115,050 29%
Labor $124,800 32%
Facilities $26,000 7%

Total $394,850

Annual Operating Costs

Maintenance $10,400
Hosting $65,000
Labor $204,750

Total $280,150
 Typical Software Costs
Startup Costs

Software Applications $310,250

Web design $1,105,000
Enterprise Integration $364,000
Other labor $403,650
Facilties $78,000

Total $2,260,900

Annual Operating Costs

Maintenance $7,150
Labor $890,500

Total $897,650
 Total Costs

Startup

Software $2,260,900 85%

Hardware $394,850 15%

Total $2,655,750

Operating

Software $897,650 76%

Hardware $280,150 24%

Total $1,177,800
 Moral of this lecture

Computer Systems can be Fast, Cheap, or Reliable

– Choose any two