Centralized processing集中処理 is a processing form in which work is gathered on a single computer (a host computer) and handled there in one place.
Users' terminals connect to the central computer, which takes on all data storage and computation. Because management and maintenance happen in one place, operations are easy to keep consistent, but if the central computer fails, the whole system stops.
AnalogyIt is like sorting every parcel in town at one large depot. Nothing escapes notice, but if that depot shuts down, all deliveries stop.
Distributed processing分散処理 is a processing form in which work is shared among multiple computers linked by a network, which cooperate to carry out the whole job.
Distributed processing
Horizontal distribution
Devices at the same level share the work
Vertical distribution
Upper- and lower-level devices take on different roles
If some machines stop, the rest can usually keep working, and capacity grows easily by adding machines; on the other hand, coordinating and managing the machines becomes more complex. When machines at the same level share the work side by side, it is called horizontal distribution水平分散, and when upper- and lower-level machines take on separate roles, it is called vertical distribution垂直分散.
AnalogyIt is like setting up a depot in each district to share the parcels. If one stops, the others keep things moving, but the depots must stay in touch with each other.
Exam points
ExamHorizontal load balancing水平負荷分散 places several computers that do the same job side by side and spreads the work so that no single machine is overloaded.
ExamGrid computingグリッドコンピューティング pools the processing power of many computers on a network and uses them as one large virtual computer.
Parallel processing並列処理 splits a series of tasks into units that can run at the same time, assigns them to multiple CPUs, and executes them simultaneously to shorten processing time.
A large job
Split into units that can run at once
Assign to multiple CPUs
Run simultaneously and combine results
The CPUs run at the same time and their results are combined, so the total time shrinks.
4 / 4
It can be done with several CPUs in one computer or shared across several computers. The less the parts have to wait for each other's results, the more time is saved.
AnalogyStuffing a pile of envelopes goes faster when several people work on it at once than when one person does them one by one.
Exam points
ExamMultithreadingマルチスレッド creates several flows of processing (threads) inside one program and advances them concurrently.
ExamScale-outスケールアウト raises the processing capacity of a whole system by adding more servers that share the work in parallel.
ExamA supercomputerスーパコンピュータ boosts its computing power through parallel processing 並列処理 on many processors and is used for scientific and engineering calculations with enormous workloads.
A client-server systemクライアントサーバシステム divides roles into the side that requests a service (the client) and the side that provides it (the server), and links them over a network.
← Scroll the figure sideways →
It works through an exchange of requests and responses: the client asks for processing, and the server sends back the result. Placing data and programs on the server lets management and updates be handled together on the server side.
AnalogyIt is like a library patron asking for a book at the counter and a librarian fetching it from the stacks. Who asks and who responds is fixed.
How to rememberA client is someone who engages a service, and a server is one who serves. The meanings of the words are the two roles themselves.
Exam points
ExamA thin clientシンクライアント gives the client only minimal functions such as input and display, and concentrates application execution and data management on the server side.
Peer-to-peerピアツーピア is a configuration with no dedicated server, in which devices on a network exchange data directly as equals.
Each device can act both as the one receiving a service and as the one providing it. Load does not pile up on any particular server, but because management is split across the devices, the whole network is hard to manage as one.
AnalogyIt is like a book club where members lend and borrow books hand to hand. There is no central desk like a library's, and everyone can be both lender and borrower.
How to rememberA peer is an equal or a companion. Read the name as peer linking directly to peer.
Exam points
ExamStandaloneスタンドアロン means using a PC by itself without connecting it to a network, which sets it apart from peer-to-peer ピアツーピア, since that only works when devices are linked over a network.
A Web systemWebシステム is a type of client-server system that uses the user's web browser as the client and delivers services through requests and responses with a web server.
Browser sends a request
Web server receives it
Server side processes it
Response appears on screen
The response returns to the browser, and the result appears on the user's screen.
4 / 4
Users can access the service with just a browser, without installing dedicated software. Features are added or fixed on the server side, so there is no need to redistribute anything to many terminals. Large systems sometimes split servers by role, for example a web server that returns screens, an application server that runs business logic, and a database server that manages data.
How to rememberA web is a spider's web. Servers spread across a worldwide web are all reached through a single entrance, the browser.
Key points
Where processing happens (centralized or distributed processing) and what role each device plays (a client-server system or peer-to-peer) are separate axes.
Parallel processing splits work into units that can proceed at once and runs them on multiple CPUs, shortening the total processing time.
A Web system is a type of client-server system that uses a web browser as the client.