MGMT 1040 Chapter Notes - Chapter 3: Multiprocessing
MGMT 1040 Chapter 3 Notes – Summary
Introduction
• Programs can be divided into independent pieces and the different parts executed
simultaneously on multiple CPUs.
• With multiprocessing, increasing computational power may be achieved by adding more
CPUs, which is relatively inexpensive.
• Data dependencies and cache memory misses can stall the pipelines in a single CPU.
• Multiprocessing allows the computer to continue instruction execution in the other
CPUs, increasing overall throughput.
• Assignment of work to the various processors is the responsibility of the operating
system.
• Work is assigned from among the programs available to be executed.
• More commonly, from independent segments of those programs called threads.
• Since each of the CPUs has access to the same memory and I/O, any CPU can
theoretically execute any thread or program currently in memory, including the
operating system.
• This raises the question of control of the system.
• There are two basic ways of configuring a multiprocessing system.
• Master-slave multiprocessing, in which one CPU, the master, manages the system, and
controls all resources and scheduling.
• Only the master may execute the operating system. Other CPUs are slaves, performing
work assigned to them by the master.
• Symmetrical multiprocessing (SMP), in which each CPU has identical access to the
operating system, and to all system resources, including memory.
• Each CPU schedules its own work, within parameters, constraints, and priorities set by
the operating system.
• In a normal SMP configuration, each of the CPUs is identical.
find more resources at oneclass.com
find more resources at oneclass.com