Disadvantages of centralized computing and the shift towards distributed/decentralized computing
With the first computer systems, centralized
computing was the only option. In centralized computing,
terminals are attached to a central computer that performs all
the computations and controls all the peripherals, such as
printers. The terminals in these systems were eventually called
dumb terminals because they could perform no processing and
served only as an input/output mechanism linking users with the
central computer.
Centralized computing to service multiple users has many
shortcomings.
First, the computer has to perform two types of work. It has to
do the computing for the users, and it has to manage the status
and progress of work being done for each online user. The
operating system software that keeps track of the jobs uses a lot
of the CPU's processing power.
As a result, a substantial percentage of the available computing
resources go toward controlling the job stream rather than
accomplishing the work the users want.
A second shortcoming is total reliance on the central computer.
If either the telecommunications line to the computer or the
computer itself goes down, no work can be done using the dumb
terminal.
A third shortcoming is the necessity to balance the computer's
workload to avoid peak-load problems when people across the
organization want to use it simultaneously. This necessitates
tight schedules and controls on the corporate computer to ensure
that it can provide reliable
service. Therefore, it limit users' flexibility in doing their
own work.
The centralized data processing shop become less important as the
numbers of departmental minicomputers and personal computers
grew.
In distributed computing, multiple workstations are linked to
share data and computing resource and use a telecommunications
network to link to other devices. The need for the network as
part of the computing system is one of the reasons for the great
attention to the convergence of distributed computing
technologies and telecommunications.
Therefore, TINA'96, the preminent research forum of the
telecommunications industry, is expanding its scope to embrace
all aspects of distributed computing, including object-oriented
systems, multimedia, distributed data management and large scale
database systems, groupware, telecommerce, distributed systems
fault-tolerance and network security.
On the other hand, decentralized processing involves completely
independent user computer systems with independent databases,
programs, applications, budgets, and information system
personnel.
Distributed computing improves coordination. It helps in sharing
data, messages and work products, and permits sharing of
resources such as printers. Data, messages, and work products are
shared in two ways.First, an individual may send a message or
file to other computers on the network.
Second, an individual may access data residing on a hard disk
attached to a different computer on the network. For example, an
insurance adjuster using a portable computer can download the
day's data to a central database. Later, an insurance pricing
analyst might access the claims database through another part of
the network. The sharing of printers is accomplished by creating
a list of documents to be printed and controlling the printing
independently of what individual users are doing on their own
computers.
The degree to which computing should be distributed depends on
tradeoffs between the cost of transmitting data over
telecommunications lines and other factors such as user
convenience, maintainability, and security. For infrequently
changed data, such as product specifications, it is usually
better to send a copy of the data to each site periodically. For
data that changes frequently, such as sales orders, it is usually
better to maintain the data in fewer sites
and use telecommunications more extensively for access to the
most current data.