1 Distributed Systems: Client/Server Computing. CS-550 (M.Soneru): Distributed Systems - Client/Server Computing: [Sta’01] 2 Client/Server Computing Major.

1

Distributed Systems: Client/Server Computing

CS-550 (M.Soneru): Distributed Systems - Client/Server Computing: [Sta’01]2

Client/Server Computing

• Major trend in information systems, replacing both the centralized approaches and alternative forms of distributed computing

• Main concepts– Clients

• Single-user PCs or workstations with user-friendly, usually graphical, interfaces

– Servers• Provide a set of shared user services to clients

• Example: Database server, controlling a relational database– Allows many clients to share access to the same database

– Can use a high-performance computer system to manage the database

– Network• Clients and servers are connected to a LAN or WAN or by an internet of

networks


Client/Server Computing (cont.)



• Client/server architecture– Platforms and the OSs may be

different– Communications protocols and the

applications supported must be common

– Allocation of application-level tasks between clients and servers is central to the architecture. It must:

• Optimize platform and network resources

• Optimize the ability of users to perform various tasks and to cooperate with one another

– Presentation services in client must provide a user interface (e.g., GUI) that is easy to use, powerful, and flexible



• Relational database applications– Server is responsible for maintaining

the database using a database management system software module

– Client has applications that make use of the database

– Client/server interactions are in the form of transactions in which the client makes a database request and receives a database response, e.g., the Structured Query Language (SQL)

– Not all the application logic (e.g., data analysis) has to reside in the client



• Example: Online access for record lookup– Server is maintaining a database of 1 million records

– User wants to perform a search using a certain search criteria (e.g., records older than 1999) and expects to find at most a few records

– First query yields a server response of 100,000 records

– User sends another query adding new qualifiers

– Response returns 1,000 records

– Client issues third request with additional qualifiers and receives a single record



• Example 2: Online access for record lookup– Server is maintaining a database of 1 million records

– Client would like to compute the grant total of a certain field across many records

– The query results in 300,000 records transmitted, which would be a misuse of the client/server architecture

– Solution: move part of the application logic to the server



• Classes of client/server applications– Wide spectrum of implementations in partitioning work between client

and server– Host-based processing

• All processing done on a central host• User’s station is limited to the role of a terminal emulator, even if it employs

a microcomputer• Traditional mainframe computing, not a true client/server architecture



• Classes of client/server applications (cont.)– Server-based processing

• Most basic class of client/server configuration, early implementations (thin client model)

• Client provides a graphical user interface• Server provides all processing• This configuration does not provide significant improvements in the

use of resources



• Classes of client/server applications (cont.)– Client-based processing

• All application processing is done by the client, except for the data validation routines

• Advantage: allows users to employ applications best suited for local needs

• Most common client-server approach in use today• Fat client model



• Classes of client/server applications (cont.)– Cooperative processing

• Distribution of application processing is optimized, taking advantage of the strengths of both client and server machines and of the best distribution of data

• More difficult to set up and maintain

• Can lead to greater network efficiency and highest productivity gains


Client/Server Computing (cont.)• Three–tier client/server architecture

– Application software is distributed between three types of machines: a (thin) client, a middle-tier server, and a backend server

– The middle-tier server provides a gateway function between the clients and the different backend servers

• Convert protocols and map from one type of database query to another

• Can merge/integrate results from different sources

• Can serve as a gateway between the new desktop applications and the backend legacy applications

– The middle-tier server is both a server and a client: server for the thin client and client for the backend server


Client/Server Computing (cont.)• File cache consistency

– To reduce the performance penalty of accessing remote files, individual systems can use file caches

– Typical distribution mechanism for caching files among network workstations: a file access request in the client results in (1) request to client cache (“file traffic”), then to local disk (“disk traffic”), then to file server (“server traffic”), cache and then disk

– Problem: caches can become inconsistent when the remote data are changed and local cache is obsolete

– Solutions: (a) File locking techniques to prevent simultaneous file access by more than one client, and (b) Sprite approach: on a write by a client, all other clients that opened the file are notified



• Middleware– The rapid development and deployment of client/server products has

outpaced the standards efforts: difficult to implement an integrated, multi-vendor client/server solution

– Developers need a set of tools that provide a uniform means and style of access to system resources across all platforms

– Middleware: • A set of standard programming interfaces and protocols that sit between the

application above and the communications software and operating system below

• Hide the complexities and disparities of different network protocols and operating systems

– Client and server vendors provide a number of middleware packages as options

– The user selects a middleware strategy and then assembles equipment from different vendors that support that strategy


Client/Server Computing (cont.)• Middleware architecture

– The role of the middleware depends on the style of client/server computing used– There is both a client and a server component of middleware– Goal: allow an application or user at the client to access different services on

servers, even though the servers’ implementations may be different– Many relational databases, even though they support Structured Query Language

(SQL), they also add proprietary extensions to SQL– Example: A distributed system supporting a personnel department is using both

Gupta and Oracle databases


Client/Server Computing (cont.)• A logical view of a distributed system using middleware

– The entire distributed system can be viewed as a set of applications and resources available to its users

– Applications run over a uniform Applications Programming Interface (API)

– The middleware operates over all client and server platforms and is responsible for routing client requests to the appropriate servers


Client/Server Computing (cont.)• An example of the use of middleware to integrate different products

1 Distributed Systems: Client/Server Computing. CS-550 (M.Soneru): Distributed Systems - Client/Server Computing: [Sta’01] 2 Client/Server Computing Major.

Documents

clientserver computing

database server

database response

client slide

database request

server slide

records user

database network clients