DISTRIBUTR OPERATING SYSTEMS

1
DISTRIBUTR OPERATING SYSTEMS

• 
  
• ANDREW S.
  
• TANENBAUM

2
Chapter 1

• INTRODUCTION TO DISTRIBUTED SYSTEMS

3

• Two advances in technology began to change the
  traditional situation
  Powerful microprocessors and high-speed local
  area networks.
• Motivation and goals of using distributed systems

4

• Advantages of distributed systems over
  centralized ones.

5

• Advantages of distributed systems over
  isolated(personal) computers.

6

• Disadvantages of distributed systems

7

• Distributed system hardware

8

• Bus-based microprocessors
  Bus traffic ? cache memory ? incoherent problem ?
  write-through cache and snoopy cache(snoopy
  write-through cache) ? 32-64 CPUs on a single
  bus

9

• Switched multiprocessors
  more than 64CPUs on a multiprocessor ? with n
  CPUs and n memories, n2 switching stages and each
  stage contains log2n switches (i.e. n/2 log2n
  switches) ? NUMA(NON Uniform Memory Access) ?
  some memory is associated with each CPU.

10

• CONCLUSION
  To summarize, bus-based multiprocessors,
  even with snoopy caches, are limited by the
  amount of bus capacity to about 64CPUs at most.
  To go beyond that requires a switching network,
  such as a crossbar switch, an omega switching
  networking, or something similar.

11

• Large crossbar switches are very expensive, and
  large omega networks are both expensive and slow.
  NUMA machines require complex algorithms for good
  software placement. The conclusion is clear
  building a large, tightly coupled, shared memory
  multiprocessors is difficult and expensive.

12

• Bus-based multicomputers
  Since it is only for CPU-to-CPU communication and
  there will be much less traffic over it, it needs
  not be a high-speed backbone bus.

13

• Switched multicomputers Two
  popular technologies are presented (a grid and a
  hypercube).

14

• Types of operating systems for the
  multiprocessors and multicomputers

15

• Network operating systems and NFS
• This is an environment with loosely-coupled
  software on loosely-coupled hardware.
• Network operating system is in a situation where
  each machine has a high degree of autonomy and
  there are few system-wide requirements.

16

• Different clients may mount the severs in
  different places. (see Fig. 1-10)
• NFS architecture
  The basic architectural
  characteristic of NFS is that servers export
  directories and clients remotely mount them.

17

• NFS protocols (two client-server protocols) The
  first handles mounting (file handle) and the
  second is for directory and file access
  (stateless).

18

• True distributed systems
• It is an environment with tightly-coupled
  software on the loosely-coupled hardware.

19

• A distributed system is one that runs on a
  collection of machines that do not have shared
  memory, yet looks to its users like a single
  computers. There must be a single, global
  interprocess communication mechanism so that any
  process can talk to any other process.

20

• As a logical consequence of having the same
  system call interface everywhere, it is normal
  that identical kernels run on all the CPUs in the
  system.

21

• Multiprocessor timesharing system
• It is an environment with tightly-coupled
  software on tightly-coupled hardware.

22

• The key characteristic of this class of system is
  the existence of a single run queue a list of
  all the processes in the system that are
  logically unblocked and ready to run.

23

• The differences the three kinds of systems

24
(No Transcript)
25

• DESIGN ISSUES
• Transparency

26

• Flexibility
• The monolithic kernel does provide the file
  system, directory system, full process
  management, or much system call handling.

27

• The microkernel just provides four minimal
  services such as IPC, mechanism, some memory
  management, low-level process management and
  scheduling and low-level input/output. It sends a
  message to the appropriate server which then does
  the work and returns the result.

28

• Reliability
• Availability refers to the fraction of time that
  the system is usable. A highly reliable system
  must be highly available, but that is not enough
  (Since inconsistent copies)

29

• Security
• Fault tolerance
• Performance degradation

30

• Performance
• A large number of small computations and highly
  interactions on a distributed system ?
  fine-grained parallelism ? coarse-grained
  parallelism ? large computations and low
  interactions.

31

• Scalability Will the methods we are currently
  developing scale to such large systems?

32

• Guiding principle avoid centralized components,
  Table, and algorithm.

33

• No machine has complete information about the
  state of the system.
• Machines make decision based only on locally
  available information.
• Failure of one machine does not ruin the
  algorithm.
• There is no implicit assumption that a global
  clock exits.