Transcript Distributed Systems - Computing and ICT in a Nutshell

Information & Communication Systems
These slides cover the following topics:
•
•
•
•
•
•
•
•
•
Centralisation of control and data – mainframes
Dispersed system – personal computers
Dispersed + networking = distributed system
Distributed storage
Distributed processing
Distributed control
Telecommunications for competitive advantage
Networking
Network infrastructure for the World Wide Web
Mainframes
• The earliest computers were
what we now call mainframe
computers
• The first commercially
available computers
appeared in the 1950s (e.g.
LEO, shown right)
• Mainframe computers centralise the processing and use
dumb terminals, which are often text only – they do no
processing, and are effectively just monitors and keyboards
• Centralised processing fell from favour in the 1980s as PCs
became cheaper and smaller
• Centralised processing is making a comeback with the
increased popularity of the internet, network computers
(NCs) and systems such as Citrix Metaframe
Centralised Processing
• Terminals are cheap
• Expansion is easy – just
add another terminal
• No software to configure
on terminals
• Centralised resources –
data and peripherals
• Fewer data
integrity/redundancy
issues
• Background processing
can continue after logout
• Security issues – data can
be accessed from
anywhere
• Space – large computer
room required!
• Permanent communication
link required
• Speed – large bandwidth
required for GUI
• Cost of maintaining link –
phone/leased line
• Limited facilities – no local
printers, disc drives, etc.
Dispersed Systems
• Everyone has their own PC – processing is
dispersed throughout the organisation
• This changed working practices, as everyone had
access to data processing facilities – e.g.
managing directors could type their own letters!
• Lack of centralisation leads to problems:
– No standardisation – e.g. shared templates
– Data redundancy/integrity problems – multiple
copies of data could be changed separately
– Communication between departments is not
possible
Distributed Systems
• When networking is added, dispersed systems
become distributed systems
• Advantages and disadvantages as per networks:
• Shared hardware,
software and data
• Easier to install
applications, etc.
• Improved
communications
• Centralised
management of users,
backups, etc.
• Dependency - server or
cable failure affects all
users (potentially!)
• Security issues passwords, restricted
access, viruses, etc.
• Performance - network
slows down under load
• … to be furthered!
Centralised Storage
• Storing data centrally has big benefits in terms of
redundancy - all users are using the same data
which can easily be kept up to date
• The DBMS needs to be careful to maintain the
integrity of the data through record-locking
• What happens if two users try to update a record
at the same time?
–
–
–
–
–
prevent access to records in use?
make open records view only?
lock open records so that they can’t be changed?
lock related records?
lock the whole table?
Client-Server
Request for information
Application
or web-site
Results of query
Client
• No redundancy
• Can connect remotely e.g. from home
• Improved security - data is
not stored by the client?
• Less traffic as only subsets
of the data are transferred
Database
Server
• The database is
centralised
• The application
or web-site itself
contains no data
• Less secure as
users may connect
remotely?
• Communication
cost for remote
connections
• Performance?
Distributed Databases
• Another approach is to distribute the database so
that all users are not connecting at the same time
over costly communication lines
• This can be done in one of three ways:
– duplicate the entire database
– split the database into sections
– create a central index to local databases
• This can cause problems:
– if the database is duplicated, how do you reconcile
changes made to one database with the other
databases?
– if someone is editing a record, how can you lock
that record or table in a database elsewhere?
Example
Head
Office
Appointment
North
Ipswich
John
• The head office
has all the data for
the whole company
East
South
Norwich
Kings
Lynn
Janet
West
Appointment
• Each region stores
only customers in
that region
• Each branch will
only store local
customers
• Each adviser only
has his/her own
client records
Head
Office
I’m off to
Kings Lynn!
North
East
South
Ipswich
Norwich
Kings
Lynn
John
West
Fred
• Such
transfers
might only
occur once a
day
Distributed Storage
• Better than dispersed
systems?
• No dependence on
central database or
server
• Faster - the data is
stored locally, so there
are no communication
bottlenecks
• Still dependent on
communications
• There are security issues
with remote connections
• There is quite a lot of
redundancy mismatches need to be
reconciled
• Inability to lock records
could cause integrity
problems
Distributed Processing
• Processing of data can also be broken down - this
is known as distributed processing…
• …or parallel processing
• This can take place on networked computers, or
computers with multiple processors
• It is best suited to processor-intensive jobs with not
much data (otherwise the overhead of transferring
the data would outweigh the benefit)
• How do you split a job?
• Examples:
– SETI (search for extra-terrestrial intelligence)
– Oxford university screen saver
Distributed Control
• Control can also be distributed, e.g. network
routing:
Source
• When the
message sets out,
there is no predetermined route
Destination
• Each node
decides the best
route, based on
local traffic, etc.
Telecomms for Competitive Advantage
• ICT is used to speed up or simplify business
processes - e.g. mail order, customer management
• For example:
– customers can order goods via a web-site, a
confirmation e-mail is sent, and the goods are
dispatched from the warehouse - no money needs to
be sent, or order forms filled in and posted.
– sometimes the goods can even be downloaded
directly from the web-site, e.g. software, video clips.
– customers can view their account activity on-line and
track orders, and there is less paperwork for the
company - they could even have automated
“pickers”
Case Study - Guardian Royal Exchange
• Old “manual” system:
– an advisor visits the customer, completes forms
outlining cover requirements and budget and
posts them to the head office
– actuaries at the head office produce the
“illustration” (quotation) and send it to the advisor
– the advisor returns to customer with the illustration
– if the customer agrees, the “proposal” form is
completed and the paperwork is sent off to the
head office
• There was a 60% rejection rate due to proposals
being completed incorrectly, or illustrations being
done for ineligible customers, etc.
Case Study - Guardian Royal Exchange
• New “computerised” method:
– The advisor visits the customer with his laptop and
enters the customers personal details
– The software suggests suitable products and
performs “illustrations”
– If the customer chooses to buy, the “proposal” is
completed on the laptop and printed out for the
customer
– Back at the office, the advisor plugs the laptop into
a phone-line and transmits the data to head office.
• Using this method, the advisor can complete the
sale in one visit, and use of cross-field validation,
etc., reduced the rejection rate to less than 2%!
Social Implications
• There can be resistance to using new technology
– this can lead to systems lying unused
– in the GRE example, so few advisors wanted to use
the laptops that they introduced a higher commission
rate for policies transmitted electronically
• The workforce will need to be trained to use new
systems
• Sometimes companies introduce new ICT systems
with the intention of reducing the workforce - in the
GRE example they would have been able to get rid
of head office actuarial staff
• See also e-commerce on the intranet
Networking
• Local Area Network (LAN) - a number of
computers on the same site, sharing a common
electrical connection
• Can share hardware,
software and data
• Easy to administer
centrally, e.g.
– users
– backups
– applications
• Improved
communication
• There is dependency
on the system - e.g.
power or cable failures
affect all users
• There are security
issues - hackers,
viruses, etc.
• Performance
degrades under load
Networking Terminology
• A gateway is used to link two different networks
together, e.g. to connect a LAN to the internet
• A bridge links together two similar networks
• A client-server network is one that has one or
more file servers storing data and applications
and print servers - this is the most common type
for anything but very small networks
• A peer-to-peer server has no server:
– files and peripherals (e.g. printers) are shared
between workstations on the network
– all workstations may need to be switched on
– suitable for small networks due to reduced cost
Client-Server or Peer-to-Peer?
• Choose client-server if:
– you have a large network that can absorb the cost
of the server
– you want centralised management of users and
resources
– users are mobile and/or don’t always have their
PCs turned on
• Choose peer-to-peer if:
– the network is small and you can’t justify the
expense of a dedicated server
– you have a workforce that generally uses the same
PC and works the same hours so that all
workstations are on at the same time
Wide Area Networks
• Some large organisations are national or multinational, and data might need to be exchanged
globally
• Large wide area networks (WANs) therefore
exist to make this data transfer:
– possible
– secure
• Examples of WANs include:
–
–
–
–
airline booking systems
credit card payment systems
national lottery machines
Police National Computer
The Internet and the World Wide Web
• The internet and the world-wide web are not the
same! The internet also supports:
–
–
–
–
–
e-mail
newsgroups
IRC chat
FTP - file transfer
telnet - for logging into other computers remotely
• Through a variety of protocols, e.g.:
– TCP/IP - transfer control protocol/internet protocol
– POP3 (post office protocol) and SMTP (simple mail
transfer protocol)
– HTTP - hyper-text transfer protocol
– FTP - file transfer protocol
Connecting to the Internet
• PSTN (public service telephone network)
– connect using a modem
– allows speeds up to 56kbps
• ISDN (integrated services digital network)
– digital phone line - dialling of number required
– 64kbps per channel - usually 2 channels = 128 kbps.
– Connect using an ISDN or terminal adapter
• ADSL (asymmetric digital subscriber line)
–
–
–
–
connect using a terminal/network adapter
“always on” - no need to dial connection
faster to download than to upload (asymmetric!)
not charged per minute (unlike PSTN and ISDN)
Internet Infrastructure
• Consumers connect to
Internet Service
Providers (ISPs), e.g.
Freeserve, AOL,
Tiscali
• ISPs connect to
Network Service
Providers (NSPs),
e.g. BT
• The links between the
NSPs is the internet
backbone
Consumers
ISP
NSP
NSP
Internet
backbone
NSP
NSP
ISP
Other ISPs
ISP