Computer System: General requirements

In general you build your computer system to satisfy a specific need. A

"computer system" is made up of : 1) hardware: Servers, PCs, networks,
storage… 2) software: operating systems and applications.

Any system should meet the following requirements:

- Reliability:
- Fault tolerant
- Scalable
- Security
- Upgrade and Update

Let’s talk in some details about each of those terms:

1. Reliability

The trustworthiness to do what the system is expected or designed to do.

Reliability metrics include the following averages:

 POFOD (probability of failure on demand): The likelihood that the

system will fail when a user requests service. A biometric authentication
device that fails to correctly identify or reject users an average of once out
of a hundred times has a POFOD of 1%.

1
 ROCOF (rate of failure occurrence): The number of unexpected events
over a particular time of operation. A firewall that crashes an average of
five times every 1,000 hours has a ROCOF of 5 per 1,000 hours.

 AVAIL (availability or uptime): The percentage of time that a system is

available for use, taking into account planned and unplanned downtime. If
a system is down an average of four hours out of 100 hours of operation,
its AVAIL is 96%. Actually we talk about the following terms: uptime, and
downtime. Uptime is the time during which a system is working without
failure. Contrast with downtime. The downtime is the time during which
a computer is not functioning due to hardware, operating system or
application program failure. In this term the AVAIL or the percentage of
availability:
AVAIL = Uptime/ (down time + uptime).
This downtime cost lot of many for example EBAY was down for
approximately three hours, the company will have to pay around $ 12
million as compensation. The reason was a crashed gateway router.

Note1: Downtime + Uptime = the measurement period.

Note2: Some times we use the term “zero downtime “ which means no
down time and from it the IBM Eserver zSeries

 Mean Time To Restore (MTTR) :is the average time taken to reinstate
a failed component to functioning state. In previous example if the first
if the first time we took 1 hor and 2 hours in the rest times, so the
MTTR = 1.8 h

 Mean Time Between Failures (MTBF) or sometime MTTF Mean Time

To Fail: mean time between failures (MTBF): An indicator of expected
system reliability calculated on a statistical basis from the known failure
rates of various components of the system. Note: MTBF is usually
expressed in hours. Of a system, over a long performance
measurement period, the measurement period divided by the number
of failures that have occurred during the measurement period.

Availability % = MTBFsys / (MTBFsys + MTTRsys) * 100

Where MTBF is Mean Time Before Failures and MTTR is Mean Time
To Repair.

2
MTBF for serial and parallel systems:

For serial systems:

For parallel systems:

Example A=0.99, B= 0.999, so the serial=? And the parallel equal= ??

3
2. Fault tolerant

To improve your system reliability you should be ready when a failure occurs.
This failure could be:

 Software failures

 Hardware failures

 Human errors

 Disasters

The fault tolerant reflects the ability to continue non-stop when a failure occurs. A
fault-tolerant system is designed from the ground up for reliability by building
multiples of all critical components, such as CPUs, memories, disks and power
supplies into the same computer. In the event one component fails, another
takes over without skipping a beat.

True fault tolerant systems with redundant hardware are the most costly because
the additional components add to the overall system cost. However, fault tolerant
systems provide the same processing capacity after a failure as before, whereas
high availability systems often provide reduced capacity.

Redundancy Multi redundant hot swappable Power supplies

There are many level of fault tolerant could be expressed by the following terms:
 hot swap : To pull out a component from a system and plug in a new one
while the main power is still on. Also called "hot plug" and "hot insertion.

4
 in Fault Tolerant Environments Hot swap is a desired feature of fault
tolerant systems built with redundant drives, circuit boards and power
supplies, especially servers that run 24/7. When a component fails and the
redundant unit takes over, the bad one can be replaced without stopping
the operation.

 hot fix : To make a repair during normal operation. It often refers to

marking sectors in poor condition as bad and remapping the data to spare
sectors. Some SCSI drives can automatically move the data in sectors
that are becoming hard to read to spare sectors without the user,
operating system or even the SCSI host adapter being aware of it.

3. Clustering

Using two or more computer systems that work together. It generally refers to
multiple servers that are linked together in order to handle variable workloads
or to provide continued operation in the event one fails. Each computer may
be a multiprocessor system itself. For example, a cluster of four computers,
each with four CPUs, would provide a total of 16 CPUs processing
simultaneously.

A cluster of servers provides fault tolerance and/or load balancing. If one server fails, one or more
additional servers are still available. Load balancing distributes the workload over multiple
systems.

4. Scalable

5
 Expandable or scalable. Referring to hardware or software, the term has
become a popular buzzword in the IT world. A "highly scalable" device or
application implies that it can handle a large increase in users, workload or
transactions without undue strain.

Scalable does not always mean that expansion is free ( like adding user to a
domain for example). Extra-cost hardware or software may be required to
achieve maximum scalability. Nevertheless, scalability is a positive feature of
a product sold to fast-growing companies.

5. Security:
(Later in chapter 14 )
6. Upgrade and Update
Usually we refer to the upgrade by replacing an existing item by a newer one.
In general, we use this term with hardware and software. While we use
update to talk about replacing existing data or file by the most recent available
one. Some vendor talk about updating when we replace an item or few while
they use upgrade when the make a hall change.

The Servers:

1. Servers Classifications:
There is no general classification used to classify the servers, this
classification refers to the vendors. The main vendors in the market today are:
IBM, HP, SUN,DELL, Fujitsu and others.
In general, and to classify the servers, we could use the following criteria:

- By model : depends to the trade mark of each vendor.

o IBM: In today’s market, IBM has five server brands:
 IBM Eserver zSeries®
 IBM Eserver iSeries™
 IBM Eserver pSeries®
 IBM Eserver xSeries
 IBM Eserver BladeCenter™
Refer to www.ibm.com for more information

6
 o HP : HP In today’s market, HP has many server brands

 HP Integrity servers
 HP Integrity NonStop servers
 HP 9000 servers
 HP AlphaServer systems
 HP e3000 servers
Refer to www.hp.com for more information

- By processors: Intel based servers, AMD based servers and others.

- By architecture: Three main architecture
o Tower or stand alone servers:
- By operating system:

Performance:
The speed with which a computer processes data. It is a combination of internal
processing speed, peripheral speeds (I/O) and the efficiency of the operating
system and other system software all working together.

 Throughput
1) In computer technology, throughput is the amount of work that a computer can
do in a given time period. Historically, throughput has been a measure of the
comparative effectiveness of large commercial computers that run many
programs concurrently. An early throughput measure was the number of batch
jobs completed in a day. More recent measures assume a more complicated
mixture of work or focus on some particular aspect of computer operation. While
"cost per million instructions per second (MIPS)" provides a basis for comparing
the cost of raw computing over time or by manufacturer, throughput theoretically
tells you how much useful work the MIPS are producing.

Another measure of computer productivity is performance, the speed with which

one or a set of batch programs run with a certain workload or how many
interactive user requests are being handled with what responsiveness. The

7
amount of time between a single interactive user request being entered and
receiving the application's response is known as response time.

A benchmark can be used to measure throughput.

2) In data transmission, throughput is the amount of data moved successfully

from one place to another in a given time period.

 Response time

According to the IBM Dictionary of Computing (which cites International

Organization for Standardization Information Technology Vocabulary as the
source), response time is:
The elapsed time between the end of an inquiry or demand on a computer
system and the beginning of a response; for example, the length of the time
between an indication of the end of an inquiry and the display of the first
character of the response at a user terminal.
There is also the concept of perceived response time, which is the time a user
senses as the beginning of input and the end of the response. It is actually
possible (though not usual) for perceived response time to be too fast (it can be
mildly disconcerting if a system responds almost instantly). However, this is not
the usual complaint.

 MIPS
The number of MIPS (million instructions per second) is a general measure of
computing performance and, by implication, the amount of work a larger
computer can do. For large servers or mainframes, MIPS is a way to measure
the cost of computing: the more MIPS delivered for the money, the better the
value. Historically, the cost of computing measured in the number of MIPS has
been reduced by half on an annual basis for a number of years.
The number of MIPS attributed to a computer is usually determined by one or
more benchmark runs

performance
Performance seems to have two meanings:
1) The speed at which a computer operates, either theoretically (for example,
using a formula for calculating Mtops - millions of theoretical instructions per
second) or by counting operations or instructions performed (for example, (MIPS)
- millions of instructions per second) during a benchmark test. The benchmark
test usually involves some combination of work that attempts to imitate the kinds
of work the computer does during actual use. Sometimes performance is
expressed for each of several different benchmarks.

8
2) The total effectiveness of a computer system, including throughput, individual
response time, and availability.