分布式系统——什么是分布式

​

Why Distribute?

• resource sharing

• physical separation

• robustness

• performance

• cost-effective system evolution

Examples

• file server, print server, etc.

• networked management

• directory and naming services

• desktop (multimedia) conferences

• large-scale computation

• etc.

Why are DSs Different?

• non-zero (and variable) message transmission time

1. 1. 
      
2. • processors executing in parallel
3. • multi-path links between them
4. • same message may arrive at different times at different processors
5. • possibility of “race” situations, i.e. nondeterminism in algorithm execution (or bugs)
6. • consistency of data at different sites
   1. 
      

• probability of partial failure of collaborating components

1. 1. 
      
2. not a concept found in single processor programs
3. • recovery from failures
4. 1. 1. 
         
   2. – use of alternative processor
   3. – migration of services
   4. – update and propagation of naming/routing information
      1. 
         
5. Fault tolerant algorithms
6. 1. 1. 
         
   2. – leadership elections
   3. – distributed synchronisation
      1. 
         
   1. 
      

• large scale

1. 1. 
      
2. • Abstraction
3. • Inheritance/object orientation
   1. 
      

• environment is not totally secure

1. 1. 
      
2. computer hosts may be protectable via architecture and OS etc.
3. • ... but link is very vulnerable to
4. 1. 1. 
         
   2. – tapping/message reply
   3. – hosts faking addresses
      1. 
         
5. • ... and hosts vulnerable to
6. 1. 1. 
         
   2. – deliberate message overload
   3. – faked services
      1. 
         
   1. 
      

Distributed Programming

• • 
    
• Remote Procedure Call
• 1. 1. 
        
  2. common approach
  3. • client-server model
  4. • caller’s arguments are marshalled by “stubs”, put in a packet and sent to remote proc.
  5. • problems with memory addresses
  6. 1. 1. 
           
     2. - need args to be sent by “value”
        1. 
           
  7. • execution semantics when failure
  8. 1. 1. 
           
     2. – at-least-once (cheap and easy but ...)
     3. – at-most-once (relatively easy)
     4. – exactly once (expensive protocol)
        1. 
           
  9. • how is server located?
  10. 1. 1. 
            
      2. – name server based on type matching
      3. – trader includes other attributes e.g. location of device, options, etc.
      4. – version management too
         1. 
            
  11. parameter security
  12. 1. 1. 
            
      2. – may need to pass encryption keys or other security tokens as parameters
         1. 
            
  13. • Interface Definition Language (IDL)
  14. 1. 1. 
            
      2. – indicates type and order of parameters
      3. – signature for matching in name server
      4. – possibly security indication for stubs
         1. 
            
      1. 
         
• Real World
• 1. 1. 
        
  2. often termed “Middleware”
  3. • support by environments such as DCE, CORBA etc.
     1. 
        
• Transparency
• 1. 1. 
        
  2. is desirable, but may not be achievable:
  3. • • 
         
     • – distribution - effects of distribution (delays etc.) should be invisible to user
     • – location - location of components should be irrelevant
     • – migration - remote objects may relocate during use
     • – failure - recovery from faults
       • 
         
     1. 
        
• Security
• 1. 1. 
        
  2. Use of cryptographic techniques to deal with:
  3. 1. 1. 
           
     2. • secrecy
     3. • authentication of individuals and messages
     4. • replay and faking attacks
     5. • can also provide access control to object methods
        1. 
           
     1. 
        
  •