Amazon and Ecommerce 2

1
Amazon and E-commerce 2

• Prof. Ian Sommerville

2
Todays lecture

• Amazon technology
• Recommendation systems
• Cloud computing

3
Amazon technology

• Technological challenges
• Maintain and manage stock of millions of items
• Handle records for 60 million active customers
• Provide rapid response to queries and orders
• Manage huge fluctuations in demand (demand
  probably increases by a factor of 10 at
  Christmas)
• Provide a recommendation system that exploits the
  large amounts of information collected from
  browsers and buyers

4
Amazon infrastructure

• Amazon provides reliability and performance by
• Using thousands of compatible commodity computers
  located in data centres around the world
• Based on open-source software (Linux) so no
  licensing fees paid to Microsoft
• Distributing data on its products, customers,
  orders, etc. across the infrastructure, using
  software trickery to ensure that copies of
  information are physically distributed
• Using so-called virtualisation technology to
  allow each physical machine to offer services to
  different clients
• Ensuring that there is sufficient capacity in the
  infrastructure to cover peak demand (Amazon sells
  computing and storage services because it has
  spare capacity)

5
Amazon service architecture
Clients
Rendering and UI services
Account services
Ordering services
Helper services
Search services
Data mining services
Storage services
6
Distribution
Ians account
Scottish DC
Ians account
Irish DC
Ians account
Belgian DC
7
Failure handling
Lookup (Ians Account)
8
Node failure
Lookup (Ians Account)
9
The Amazon fulfilment system

• Watch a video taken in an Amazon warehouse
• http//news.bbc.co.uk/player/nol/newsid_7130000/ne
  wsid_7136300/7136393.stm?bwbbmprmasb1news1
  ms320ms_javascripttruebbcws2
• This demonstrates the scale of Amazons
  operations, how bar coding is fundamental for
  selecting and tracking goods and the mechanical
  as well as the computer systems used by Amazon

10
Recommendation systems

• Retail commerce actually encompasses two distinct
  activities
• Searching. You know what you want and you search
  to find it. In a store, this may involve going to
  a certain department, asking for a product, etc.
  In an e-commerce site, it usually involves using
  a search engine.
• Browsing. You dont have a clear idea what you
  want but you are interested in knowing whats
  available. You want to look around a store to see
  if they have something that you might want to
  buy.
• Stores increase their business by turning
  browsers into buyers. Their floor space is often
  organised to make browsing easy.

11
E-commerce systems

• Searching is easier than in a store. If you know
  what you want, then you can find it and buy it
  quickly.
• Browsing is much harder. You cant wander around
  as you can in a physical store and the limited
  window means that you cant use the physical
  organisation of the space to attract viewers
• This is where recommender systems come in. These
  systems are designed to suggest items that
  potential buyers might be interested in. If they
  get this right, then the viewer is more likely to
  become a buyer.

12
Recommender approaches

• Make recommendations based on users past
  behaviour
• Make recommendations based on the behaviour of
  other users who have behaved in a comparable way
• Make recommendations based on the item being
  purchased
• Make recommendations based on user similarity
• Most recommender systems use combinations of
  these. Amazon mostly uses the first three items.

13
Amazon example
14
Amazon example 2
15
How does Amazon create recommendations?

• We dont know for sure - this is a business
  secret. Recommendations are one of the things
  that give Amazon an advantage over its
  competitors
• Two key elements to the system
• Linking. There are a huge number of links that
  are maintained between items in the system and
  customers. These track whats been bought, whats
  been viewed, etc.
• Data mining. Programs process all the links in
  the system and build tables that can be quickly
  accessed and displayed.
• So, for example, the data mining system can
  discover that X of people who bought item X also
  bought item Y.

16
Linked structures
Dead Souls
iGroove speakers
bought
Ian Sommerville
Sandisk SD Card
Looked at
Canon Ixus
Nikon Coolpix
Is a
Compact digital camera
Reviews
Joe Jones
Sarah Smith
Fuji Finepix
17
Class exercise 2

• Use the account that you created for the tutorial
  (or any other account) to log in to Amazon
• Choose a category that you havent looked at
  before (e.g. I chose DIY and Tools)
• Click on something (doesnt matter what)
• Scroll down the page and look at the
  recommendations. Get a picture of the links
  created
• Look at Customers also bought. Scroll (if
  necessary) and click on a book
• Look at the recommendations - what do you notice?
  What does this tell you about the recommendation
  system?

18
Cloud computing

• Amazons infrastructure is provided on a cloud
  of hundreds of thousands of computers located in
  huge data centres in different locations.
• As well as books, electronics and other goods,
  Amazon also sells computer time and storage on
  its cloud.
• This general idea of cloud computing, relies on
  organizing systems as a set of services, which
  can be delivered to multiple clients, who pay for
  services as required.
• Services are capabilities that are owned by a
  service provider but offered to clients.
• E.g. You dont normally have a personal
  hairdresser but use a hairdressing service

19
Everything as a service

• Software as a service
• Instead of software running as separate programs
  on your own computer, the software runs on the
  cloud and is accessed through a web browser.
  You dont know or care where the system is
  physically located e.g. Google mail or Zoho.
• Platform as a service
• Instead of writing programs to implement your
  web-based application on a separate server or PC,
  you write these programs using a cloud platform
  e.g. Google AppEngine or MS Azure.
• Infrastructure as a service
• Instead of buying computers and storage
  capability, you rent what you need, when you need
  it from an external provider such as Amazon.

20
Benefits of cloud computing

• Software as a service
• Software can be accessed from any computer or
  device, wherever there is an internet connection.
• No need to install software, pay for upgrades or
  pay staff to support systems.
• Platform as a service
• Cloud platforms are designed for scaleability so
  that as demand increases, the service offered by
  your system will not be degraded. This makes it
  much easier to develop large-scale web-based
  systems.
• Infrastructure as a service
• You dont have to make up-front capital
  investments in computers and storage systems.
  Rather, you simply rent what you need, when you
  need it from a cloud provider. It is very easy
  and quick to increase and decrease capacity.
• The cloud provider takes care of issues such as
  backup and recovery from failure. You dont need
  to spend money on redundant hardware that may
  never be used.

21
Impact of cloud computing

• Individuals
• Reduced costs less need to buy separate
  applications and new hardware
• Access to data from multiple devices (computer,
  phone, etc.)
• Easier sharing of information
• Businesses
• Reduced costs for hardware and people to install
  and maintain systems
• Easier to support a distributed workforce/home
  working etc.
• Lower CO2 emissions because of more efficient
  utilisation of hardware
• Society
• Easier access to computation capability for
  citizens
• Environmental benefits
• Stimulates new knowledge-based business
  development

22
Summary

• Amazon technology
• Built around the notion of cloud computing where
  thousands of cheap computers provide the platform
  for the application system
• Designed for reliability and scaleability. Always
  available, should not lose user information,
  should scale to handle large increases in demand
  without significant loss of service
• Recommender systems
• Designed to replicate the browsing experience of
  shoppers
• Rely on capturing a vast body of information
  about users, their buying and browsing habits
• Cloud computing
• Amazon also sells computing services in the
  cloud. These allow people to access capabilities
  over the internet without installing their own
  hardware and software.