[OS] Threads 1

Threads Part 1

Motivation for threads

• Problem
  • A single process cannot benefit from multi-cores
  • Very cucmberome to write a program with many cooperating processes
  • Expensive to create a new process
  • High communication overheads between processes
  • Expensive contect swithching between processes
• What if separate out a process’s execution state from others
  • Process: address space, resources, other general process attributes(e.g. user ID, privileges, …)
  • Execution state: PC, SP, registers, etc…
• Such an execution state is usually called a thread

Thread

• A thread executes a sequence of instructions in a program
• Single-threaded process
  • has one program counter specifying location of next instrution to execute
  • process executes instructions sequentially, one at a time, until completion
• Multi-threaded process
  • has one program counter per thread
• Threads share an address space
  • Code
  • Global variables
  • Heap
  • Opened files
• A thread has its own
  • Set of registers including PC & SP
  • Stack
  • Thread ID
  • TCB (Thread Control Block) : unit of a context switch, smaller and cheaper than PCB

Address space with Threads

Process vs Thread

• A process can have multiple threads
• A thread is bound to a single process
• Processes are containers in which threads execute -PID, address space, user and group ID, open file descriptors, current working directory, etc…
• All threads see the same address space
• Threads are the unit of scheduling

Multithreaded Server Architecture

Benefits of Multi-threading

• Creating concurrency is cheap
• Allow to fully utilize multi-core architectures
• Resource sharing become obvious (Threads share the address space)
• Improves program structure (can divide large task across several cooperative threads)
• Throughput (by overlapping computation with I/O operations)
• Responsiveness (can handle concurrent events)

Concurrency vs Parallelism

Types of Parallelism

• Data parallelism
  • Distributes subsets of the same data across multiple cores, same operation on each
• Task parallelism
  • Distributing threads across cores, each thread performing unique operation

Amdahl’s Law

• Give performance gains in latency from adding additional system resources to an appliation
• When we add more cores to a process (S: serial portion, N: processing cores), then
• Serial portion of an application has dispropotionate effect on performance gained by adding additional cores
  • e.g. 25% serial, 75% parallel, increasing cores from 1 to 2 results in speedup of 1.625 times
  • As N approaches infinity, speedup approaches 1 / S

Concept Review

User Threads and Kernel Threads

• User Threads : management done by user-level threads library, supported above the kernel(managed without kernel support)
  • POSIX Pthreads
  • Windows threads
  • JAVA threads
• Kernel Threads : supported and managed directly by the operting system
  • the unit of scheduling
• A relationship must exist between user and kernel threads
• Thread libraries can provide various user-level thread models on top of kernel threads

Thread libraries

• provides programmer with API for creating and managing threads
• Two primary ways of implementing
  • Library entirely in user space
  • Kernel-level library supported by the OS

Pthreads

• A POSIX standard API for thread creation and synchronization
• provided either as user-level or kernel-level
• API specifies the behavior of the thread library
• Implementation is up to development of the library (specification not implementation)

Pthreads example