Notes On Chapter Twenty-Eight -- Client-Server Interaction

• 28.1 Introduction
  
  
  • This section (group of chapters) covers network applications.
    
    
  • It covers principles, techniques, programming interface, and the structure of the software.
    
    
  • This particular chapter is about the client server paradigm. The nature of networks engenders the client server paradigm.
    
    
  • The chapter presents examples
    
    
• 28.2 The Functionality Application Software Provides
  
  
  • Applications are what users really care about: programs that send and receive e-mail, browse the WWW, games, etc.
    
    
  • It is important that applications allow users to deal with information using the language and symbols that are most natural and convenient.
    
    
  • Applications define names for hosts and services.
    
    
• 28.3 The Functionality An Internet Provides
  
  
  • Internet service is similar to telephone service in that it provides the means to communicate. However it is the responsibility of a pair of processes to establish contact.
    
    
  • The protocol software does not provide services to clients or initiate contact with processes. These are roles of application software.
    
    
• 28.4 Making Contact (The raison d'etre of the client-server paradigm)
  
  
  • Interestingly, internet service does not have the analogue of a telephone bell, and remarkably that makes a big difference in determining what style of communication tends to work well in this medium.
    
    
  • QUESTION: What works well? ANSWER: An application (a server) informs the protocol ahead of time that a certain form of message is desired and expected.
    
    
• 28.5 The Client-Server Paradigm
  
  
  • The passive listener is the server and the active initiator of communication is the client.
    
    
• 28.6 Characteristics Of Clients And Servers
  
  
  • The typical client:
    • arbitrary process, temporary client
    • invoked by user
    • executes one-shot service request
    • runs on the user's PC
    • initiates contact with server
    • contacts one server at a time
    • does not require special hardware or OS
    
    
  • The typical server:
    • special purpose
    • privileged process
    • dedicated to providing one service to multiple clients concurrently
    • starts up at boot-time
    • executes continually
    • runs on a special shared computer
    • waits passively for client contact
    • requires special hardware
    • requires sophisticated OS
  
  
• 28.7 Server Programs And Server-Class Computers
  
  
  • The server is the process that waits, offering a service. The host (machine) on which the server process runs is not the server. It's the server machine. A server-class computer is a powerful computer suitable for use to run server software.
    
    
• 28.8 Requests, Responses, And Directions Of Data Flow
  
  
  • Client-server interaction is not limited just to a simple request/reply format.
    
    
  • Large amounts of information can and do pass between client and server in both directions.
    
    
• 28.9 Transport Protocols And Client-Server Interaction
  
  
  • Both client and server use a transport service like TCP/IP to communicate.
    
    
  • TCP can be the virtual network used by applications.
    
    
• 28.10 Multiple Services On One Computer
  
  
  • It's quite simple for a computer with a multiprogramming operating system to run multiple clients and servers over a single physical network connection.
    
    
  • Often it is practical and cost effective to run many servers on a single host. Hardware is shared and excess traffic may not be a problem.
    
    
• 28.11 Identifying A Particular Service
  
  
  • There are 16 bit port numbers (64K of port numbers) to identify services.
    
    
  • We say a server "waits on a port number" or "listens on a port number." That means it specifies the service it offers by registering a service number (port number) with the protocol software.
    
    
  • In its request message, the client includes the port number of the desired service.
    
    
  • Software on the server uses the port number to "route" the message to the appropriate server -- the one "listening" on that port number.
    
    
• 28.12 Multiple Copies Of A Server For A Single Service
  
  
  • To have lower average wait time for clients, it is usually desirable to allow several clients to obtain service concurrently. The effect is much as if the clients take turns sending packets.
    
    
  • For example, if you have 12 combines it is better to let them do work for 12 different farmers than to have all 12 combines work for the same farmer at any given time. Even though the throughput of the combines is about the same either way, if one farmer has a huge amount of crop to bring in, all the other farmers may have to wait for the entire time that it takes to bring in the large harvest.
    
    
• 28.13 Dynamic Server Creation
  
  
  • Often there is a main server thread that does the waiting and spawns service threads to service one client request.
    
    A service thread may be a "one-shot" - created just to serve one client and then "die".
    
    
• 28.14 Transport Protocols And Unambiguous Communication
  
  
  • The "full" address of a client consists of the combination of its IP address and port number.
    
    
  • Similarly the "full" address of a server thread consists of the combination of its IP address and port number.
    
    
  • The protocol software can direct client packets to the correct server thread based on the full client and server addresses.
    
    
• 28.15 Connection-Oriented And Connectionless Transport
  
  
  • User Datagram Protocol (UDP) is a member of the TCP/IP suite. UDP is a tool for doing connectionless communication.
    
    
  • TCP is connection-oriented.
    
    
  • Programmers create some client-server applications that use TCP to communicate. Others use UDP. Some are able to use either or both.
    
    
  • When using connectionless transport an application must send the destination address with each message. For one-shot or small sequences of messages this may involve less overhead than connection setup would.
    
    
• 28.16 A Service Reachable Through Multiple Protocols
  
  
  • The same service may be made available to clients over two or more transport protocols.
    
    
  • This multi-protocol service might be implemented with separate servers for separate transport protocols, or a single server may be willing to "speak" one of many protocols to a client -- like a bilingual salesperson in a department store.
    
    
• 28.17 Complex Client-Server Interactions
  
  
  • Client-server interaction is flexible: A process or thread is not locked into one role.
    
    
  • For example, clients can request different services at different times from any server that offers the service.
    
    
  • A server S with client C can become a client of another server in order to obtain part of the data that C has requested.
    
    
  • We can set up a network with lots of clients and servers and let them interact in many different ways -- like a huge flea market.
    
    
• 28.18 Interactions And Circular Dependencies
  
  
  • Unfortunately, circular dependencies can arise when servers can act as clients of other servers. (The woman who sells spices goes to the shop of the man who sells soup and orders spicy soup. This causes the man to go to the woman's shop to buy spice. Each waits for the other to return.)
    
    
• 28.19 Summary