In a recent article on the new politics of the internet:
We don't offer a ready-made program, but an entire operating system. — Marina Weisbad Pirate Party (Germany)
We see operating systems taking on different meanings as it becomes a more popular term in today's cultures. What do operating systems mean to us?
2. What's a System?
Oxford English Dictionary (original 1928 version) defines a system as:
An organized or connected group of objects.
A set of principles, etc.; a scheme, method
Origin: from Greek systema σύστημα meaning organized whole, government, constitution, flock of birds, musical interval, verses in a poem
Its roots mean "set up with".
3. What's an Operating System?
Encarta defines an operating system as a master control program (MCP) in a computer.
The MCP was the first operating system that ran on the Univac I.
Trivia: MCP was also the computer program villain in the 1982 Disney film Tron.
Wikipedia v531774181 (2013-01-07) defines an operating system as a collection of software that manages computer hardware resources and provides common services for computer programs.
This definition is more precise, talking about the management (control) of resources such as diskspace, CPU cycles, and RAM and provision of common services such as network services.
Our textbook Principles of Computer System Design: An Introduction gives this definition (Section 1.1.2 page 8):
A system is a set of components that has a specified behavior observed at the interface with its environment.
Here's one way to picture a system:
But in real life, systems aren't usually this simple. You can split up systems in many different ways and these different splitups can interact differently. This becomes a design issue aimed at reducing complexity, something the textbook discusses in more detail later on.
The book gives an example of choosing from several different granularities (Section 1.1.2 page 9). An aircraft as a flying object has body, wings, and engines as components, the atmosphere and earth as the environment, and gravity, engine thrust, and air drag as the interfaces.
On the other hand, an aircraft as a passenger-handling system has seats, flight attendants and air conditioning system as components, the set of passengers as the environment, and the softness of the seats, the meals, and air flow as the interfaces.
Furthermore, subsystems are defined as systems which in one context are components in another. So we see a recursive characteristic for systems. Techniques such as modularity, abstraction, layering, hierarchy, and iteration help cope with complexity by taking advantage of this inherent recursion in systems.
4. Problems with Computer Systems
There are 4 major problems with computer systems:
Tradeoffs (Section 1.1.1.4 page 6)
Incommensurate Scaling (Section 1.1.1.3 page 5)
Emergent Properties (Section 1.1.1.1 page 4)
Propagation of Effects (Section 1.1.1.2 page 4)
Tradeoffs
When optimizing for certain features, you run into what is known as the waterbed effect.
Pushing down on one end causes the other end to pop up.
Example 1: Sorting
Let's say we have the following:
10GB of data
512 byte records
With a little math, this turns out to be roughly 220 records.
Imagine we are trying to sort this data using an algorithm like mergesort, which runs in O(nlogn) time.
Problem: With 220 records, this is going to take an eternity to copy the data and whatnot!
Instead, the client wants a solution that is 100x faster.
Solution: Instead of copying data, use pointers.
As a result, instead of copying an entire 512 byte record, we only have to copy 4 bytes for the pointer!
512/4 = ~128x speedup!
Downsides:
Deferencing pointers takes time
Memory overhead of using pointers (1/128th the size of data)
Example 2: RSA Secure ID
The RSA Secure ID is a token that adds another method of authentication for a user to a network resource. It is intended to add an extra layer of security to password authentication. However, the adoption of the RSA Secure ID has been recieved with mixed feelings.
Pros: Increased security by having another form of authentication. Cons: The trouble of carrying the ID around, price, etc.
Incommensurate Scaling
What does it mean? Not everything (normally quantitative attributes) scales at the same rate. So a solution that works for you now may nottwork when it is scaled up or down.
Example 1: The Human Body
Why are we only 2 meters tall, and not 20 meters?
The issue is that as sometting grows larger, surface area
increases at a rate of n2,
while volume increases at a rate of
n3.
Thus, our bone strength, which is a function of cross sectional area (n2),
grows much slower than our weight (n3), making 20 meter tall humans infeasible. The solution is to become like Wolverine, and develop adamantium-based bone structure.
Example 2: Economies of Scale • A Wealth of Nations by Adam Smith
A large pin factory producing for everyone is better than everyone producing their own pins. When scaling up, producing additional units becomes cheaper.
On the other hand, creating a 512-port ethernet switch is much more expensive than an 8-port switch. This is known as diseconomies of scale, where scaling up any higher is actually more expensive.
Emergent Properties
Properties (normally qualitative), usually unanticipated, that arise once all parts of a system have been assembled together.
Example 1: Internet on Campus
Internet around campus was meant to be used for learning activities, but instead was used for:
Napster (or other illegal pirating activities)
Browsing of adult material on the web
Computer viruses and worms spread through networks instead of traditional methods of infecting floppy disks
And who knows what else.
Example 2: Tacoma Narrows Bridge
No one expected the wind to match the resonant frequencies of the bridge.
Since such catastrophe never accounted for in models, this event was never considered.
Only after the bridge was assembled (and later destroyed) did architects realize this was a property which emerged from scaling up the model bridges.
Propagation of Effects (PoE)
Something that happens now will affect something else in the future.
Example 1: Reagan Becomes President
Back in the days before Reagan's presidency, he had acted in a movie, and became well known.
He obtained the role because the previous actor had committed suicide.
Through propagation of effects, Reagan indirectly became president because of a suicide.
Example 2: PoE in Engineering
Consider the location
\usr\bin\foo
Now consider
\usr\bin\涼宮ハルヒ.
If we were the computer, and we did not know the encoding of this text,
what could possibly happen is that the Japanese characters can contain the exact bits which match that of a backslash.
Thus, the computer would see the following:
\usr\bin\?\???
As a result, the computer would output an error saying the folder does not exist.
Solution: Originally the problem was solved by taking into account the type of encoding. Now non-ascii characters are encoded using UTF-8.
Propagation of effects occur all the time in chaotic systems, but it can also happen in an engineered system. Because propagation can also occur in a controlled environment, it is important to design the system effectively to localize damages that may otherwise propagate throughout.
