Lecture 4: System Architecture and Multiprocessing Systems

1. Computer System Architecture

• Single Processor System: Uses one general-purpose CPU for all OS and user instructions. Performance is limited.
• Special-Purpose Processors: Additional processors (like GPU, I/O processor) that handle specific, heavy tasks under the main CPU's control.

2. Multiprocessor Systems (Parallel/Tightly-coupled)

• Definition: Systems with two or more CPUs that share main memory and work together.
• Advantages:
  • Increased Throughput (more instructions executed per second).
  • Economy of Scale (cheaper to share resources).
  • Increased Reliability (Fault Tolerance - system degrades gracefully if one CPU fails).
• Types of Multiprocessing:
  • Symmetric Multiprocessing (SMP): All processors are peers, share memory, and can perform any task. The OS schedules all equally. Offers better scalability and balanced workload.
  • Asymmetric Multiprocessing (AMP): Each processor has a specific role (Master/Slave). One master controls the others. Easier design but less flexible.
• Dual-Core Design: Two independent CPU cores located on a single physical chip, executing instructions simultaneously.

3. OS Structure and Multitasking

• Multiprogramming: Keeping multiple jobs in memory so the CPU always has something to execute while other jobs wait for I/O, maximizing CPU utilization.
• Timesharing (Multitasking): An extension of multiprogramming that switches the CPU quickly between many users' programs, giving the illusion of simultaneous execution (quick response time).

4. Clustered Systems

• Definition: Multiple whole computer systems (nodes) working together, usually sharing storage via a Storage-Area Network (SAN). Provides high availability and fault tolerance.
• Types of Clustering:
  • Symmetric Clustering: All nodes are active simultaneously, sharing the workload and monitoring each other for failures. (More complex, more efficient).
  • Asymmetric Clustering: One computer is active, and the other is on standby (hot backup). The standby only takes over if the main node fails.
• Distributed Lock Manager (DLM): Ensures that two nodes do not modify the same file or data at the same time, maintaining consistency.

Audio Summary