|
Servers
have evolved tremendously over the last 30 years. The
evolution is influenced by changing business requirements
and applications rather than technological developments.
by Minu Sirsalewala
The
Indian enterprise has witnessed the evolution of servers
over a span of 30 years (See box: From mainframes to
blades). From mammoth 'iron boxes' hogging entire floor
levels, to sleek Web servers in the form of racks and
blades, servers have indeed evolved tremendously. For
businesses, servers form the crux of its IT infrastructure.
And server technology is constantly evolving to keep
up with ever-changing business requirements. Enterprises
have deployed many servers over the years for the various
applications that are so crucial to business. The Network
Magazine team took a trip down memory lane and explored
how servers have evolved in Indian enterprises from
the Petroleum, Banking and Manufacturing sectors. Our
first stop was HPCL where we spoke to an enthusiastic
IS manager.
Harsh Kumar, Advisor-IT, HPCL took us back to the era
when HPCL first deployed 'big iron' at its office. He
explained how these "boxes" have evolved over
the years.
"HPCL
was amongst the first few Indian companies to deploy
servers for commercial use," claims Kumar. "Others
like the Indian Railways, a few airline and insurance
companies also made use of servers then. Way back in
the late 60s we used the 1401 (IBM mainframe) that ran
on Autocoder. This machine was robust and was used for
15 years."
BUSINESS NEED
Harsh Kumar added that in the 60s and 70s a program
was generally static—once written it wasn't modified
for a long time. This is because the business needs
and processes were not dynamic and did not change frequently.
Thus a particular application could run for years without
any upgrade.
The reason why servers have undergone changes over the
years is directly proportional to ever-changing and
growing business needs. Server requirements are directly
related to the expanding business and business processes.
"With
the need to grow (business growth) comes the need for
applications, and these applications in turn raise the
need for a particular server to execute the application,"
says Kumar.
Then HPCL moved to the next era when the corporation
went in for a Distributed Data Processing (DDP) system.
This was when the client-server model became widespread.
"This
system required a small server at every location (around
200 locations) across the country connected to the main
server at the central location. These boxes ran on SCO
Unix, whereas the others were Windows NT with Windows
servers," says Kumar.
These machines were not the so-called 'server-class'
machines as they were not redundant and did not have
robust architecture.
In the 90s, HPCL deployed HP 9000 with Cobol as the
programming language for its business applications.
These were highly robust machines and were used for
eight to nine years. HPCL recently phased out these
machines due to change in applications.
The current set-up at HPCL includes a mix of various
boxes from a number of vendors specific to the applications
running in the corporation. It also has Intel Xeon 4-way
machines running Windows 2000. Kumar pointed out that
servers in his corporation have evolved over a period
as per changing business environments and needs.
DYNAMIC INFRASTRUCTURE
Our next stop was a financial institution that invests
heavily in server infrastructure, that's accommodated
on three floors of its own data center. Unlike HPCL,
this institution does not have a long history of servers.
However, it does have a modern and sophisticated set-up
that undergoes frequent changes. We met an obliging
VP-IT at HDFC Bank.
"We
do not have a systems history dating back to the 60s
or 70s—ours is a more recent set-up as the bank
was set up only in the 90s," says S.R. Balasubramanian,
VP-IT at HDFC Bank. "But we do have a complex set-up
that has undergone major changes with the changing business
processes and expanding network across the country."
Bala informs that the bank had a centralized data center
from day one, when it used Intel servers on SCO Unix
for banking applications. Initially the bank had two
branches, but as more branches were introduced, it slowly
moved to the Unix platform. This clearly indicates the
change based on new applications.
According to Bala, another factor that influenced the
shift was the latest technology available, keeping in
mind the reusability of old servers for test environment
and less resource intensive applications. Bala added
that the cost of hardware has come down drastically
over the years in India.
"The
Intel server which used to cost Rs 6 lakh five years
back is no more available. In its place there are 10
times faster servers (clock speed), with 15 times more
RAM capacity; 10 times higher disk capacity, and with
redundant power supply (RPS) at almost two-thirds the
price.
HDFC Bank is using rack servers as they take up less
floor space. It can stack up to 10 servers in a single
19-inch rack. The bank is also using blade servers.
It is possible to fit 10 or 12 blade servers in a single
4U space in the rack. This provides for as many as 80
to 90 servers in a single 42 U rack, informs Bala.
Bala explains that it is the buyers market. The vendor
who understands the buyer's requirement can be ready
with a product that is rich in features (self diagnostic
tools with self healing software, remote monitoring
tools etc.), at an affordable price.
The price of high-end machines (Unix servers) has also
come down in the recent years with more features, higher
speed, CPUs, RAM and disk capacity.
"Huge
investments are made in deploying these boxes. In the
dynamic environment we operate in, there are applications
and processes implemented every hour. This means getting
in new boxes to execute these applications, or upgrade
the existing systems," says Bala.
And what does HDFC do with servers that are detached
from core operations? It uses the old servers for running
less intensive applications or as fall-back servers
so as to ensure minimum downtime.
APPLICATION DRIVEN
From the Banking sector we traverse to Manufacturing.
We chose Honda Siel Cars India, and discovered that
this company too does not have a long history of servers.
Hilal Isar Khan, Head-IT Honda Siel Cars India Ltd,
says they started out in 1997 with basic infrastructure
comprising a UTP and hub-based network catering to the
basic transaction systems (the proprietary ERP of Honda
Motors Japan called Honda Integrated Package or HIPACK).
"This
application was hosted on IBM AS/400 using DB2 as the
RDBMS and RPG and CL as the programming languages. When
we felt the need for a mailing system, we deployed IBM
Netfinity servers—Windows NT based servers with
Lotus Notes for messaging. This clearly indicates that,
as and when applications have been needed, the servers
follow suit to execute these applications smoothly and
effectively," says Khan.
According to Khan, with the expansion of operations
and addition of applications like financial accounting,
payroll, HR, appraisals, training etc, new systems had
to be deployed. Honda Siel also deployed some IBM x-series
servers with Windows NT when it expanded its network,
communication system, and set up backup and disaster
recovery systems. Today Honda Siel has different servers
for different type of applications—and it's all
need based of course.
| Sizing
for data processing |
|
The
nature of applications performed by a computer
is another aspect of its size. In general, larger
computers are used for a broader range of applications,
than are smaller computers. That is because the
largest computers support more users, with more
diverse needs, than do smaller computers. To illustrate
this, consider a typical bank. Banking executives
use microcomputers to do 'what-if' analysis, make
decisions, etc. Each branch may have a minicomputer
to support a variety of needs for the individual
branch. And the bank may have a centralized mainframe
computer that supports all the bank's branches,
providing for an even broader range of needs.
In other words, the bank uses microcomputers for
applications at the individual level, minicomputers
at the departmental or branch level, and mainframe
computers at the corporate level. In a distributed
multi-branch banking system, microcomputers are
used by a few people at each branch (mainly people
who have decision-making responsibilities) to
do non-routine tasks. Minicomputers used at the
branches are responsible for local processing
services such as savings and current accounts,
and loans. Each machine uses a copy of the same
software for these functions. The mainframe at
the headquarters performs all the consolidated
accounting functions and is responsible for processing
both transactions through the bank's network of
ATMs and credit card transactions. Sometimes,
computers of various sizes have similar applications
that vary only in the volume of data processed.
For example a corner retail store may use a microcomputer
to manage its inventory management and a large
chain of retail stores may use a central mainframe
computer for inventory management. All three stores
use their computers for essentially the same function—it
is the volume of data processed that distinguishes
them.
Source:
IBM Mainframe Handbook by Alexis Leon
|
| From
mainframes to blades … |
|
It
all began with the mighty mainframe in the 50s
and 60s. Then came minicomputers in the late 60s
and 70s, followed by personal computers and servers
from the 80s to present day. Here's a brief overview
of each generation of servers.
Mainframes
These are very large and expensive computers capable
of supporting hundreds, or even thousands, of
users simultaneously. In the hierarchy that starts
with a simple microprocessor at the bottom and
moves to supercomputers at the top, mainframes
are just below supercomputers. In some ways, mainframes
are more powerful than supercomputers because
they support more simultaneous programs. A mainframe
can process several million-program instructions
per second. Mainframes are mainly used for voluminous
online transaction processing activities, typically
seen in insurance companies, banks, and the airline
and railway reservation systems. IBM is an established
maker of mainframes and some of their popular
models are 3090, ES/9000, S/390, and Z800. IBM
has now reclassified its mainframes as the z-Series
of servers.
Minicomputers/midrange
Also known as mid-range computers, these were
first developed as special-purpose mainframe computers.
They were used, for instance, to control machines
in a manufacturing unit. They are now widely used
as general-purpose computers. Thus the line between
minis and mainframes is constantly blurring. Minicomputers
work well in a Distributed Data Processing (DDP)
environment. That is, the organization's processing
power is decentralized, or distributed across
different computers. An example is the client-server
model, in which end-users can process at their
own microcomputers. Minicomputers are used when
large groups need access to data simultaneously.
The minicomputer can do this because the hardware
is designed for plugging in more devices, and
the CPU and support chips are designed for heavy
workloads. An important measure for mini and mainframes
is the reliability of the machine as it generally
has to operate 24 hours, where every minute of
operation is important to the company. Downtime
translates to business loss for the organization.
Popular makers of minicomputers include DEC (Digital
Equipment Corporation) that built the VAX minicomputer
used in universities, banks, and engineering firms.
IBM also creates mini computer range with the
branding of AS/400. HP has a range branded HP9000.
Microcomputers/PC
They are small, single user systems that provide
a simple processor and just a few input/output
devices. It is also known as the PC. The category
is termed 'micro' because the fundamental component
that allowed the categories development was the
CPU created on a single chip (this technology
was affordable in comparison to the mini/mainframe)
created with the introduction of 'microprocessors.'
The microcomputer was often used by a single person
for single activities. They are used by medium-sized
business and small sites of larger organizations.
They are also used in factories to control manufacturing,
process control, etc. where numerous equipment
have to be coordinated.
Rack & blade
In an IDC where 100s of servers are housed, there
is a need for servers that can be stacked in limited
space, and also the computing power used should
be low. These requirements led to new categories
of servers like rack and blade. Rack servers have
sleek, space saving designs, and are stacked in
a rack. They are ideal for jobs that need plenty
of performance in a compact, self-contained unit
but don't require back-office "big iron servers"
and are mainly used in data centers and in buildings
where space is a limitation. The height of these
servers is designated as 1U, 2U etc (1U= 1.75
inches, 2U = 3.5 inches etc). HP/Compaq, IBM,
Sun, Dell, Acer, NEC are the popular manufacturers
in this category. Blade/ultra-dense servers are
designed for environments where both space and
power consumption are constraints (like data centers).
So they are an improvement over rack servers (from
the space aspect). They are like cards, each having
their own processor, chipset and memory, but sharing
a common power supply and storage resources. They
use low-voltage processors like Pentium III 700
MHz. As they are stacked like PC adapter cards,
they save on space—you can fit more servers
per tower box. They work well for dedicated applications
like mail servers, media streaming, Web hosting.
etc.
|
Minu
Sirsalewala can be reached at minus@networkmagazineindia.com
|
|| Feedback
|| Subscribe-
