2022-04-24
  • Address space
    • Virtual memory
    • Program is "alive": OS allocates resource for a program
    • Stack vs. Heap
  • Process vs. Thread
    • Reason for thread: speed up (context switching)
    • Reason for process: split memory
    • Balancing: process for "big" code and thread for "small" code
    • Which to use: depends on the problem, not computer
    • Co-design: design coupled but fast system
    • Never enforce instruction order that does not actually exists (necessary)
  • Client/Server with threads
    • Threads within clients: RMI as an example
  • Multi-threaded server architectures
    • Worker pool: fixed pool of worker threads; inflexible
    • thread per request/connection/client
    • Thread-per-Request: Max throughput, high overhead over creation & destruction
    • Thread-per-Connection: lower overhead, unbalanced load
    • Thread-per-Object
    • Server threads and throughput