⚡ Concurrency, Parallelism & Scheduling
Threads, locks, async, schedulers the hardest problems in computing after naming things.
→ See also: OS & Kernel · Systems Track · C++
🧵 Concurrency vs Parallelism
Concurrency: Dealing with multiple things at once (structure)
Parallelism: Doing multiple things at once (execution)
Concurrency is about composition. Parallelism is about speed.
Rob Pike
📚 Courses
| Course |
Platform |
Link |
| Concurrent Programming in Java |
Coursera (audit, Rice) |
Coursera |
| Parallel Programming in Java |
Coursera (audit, Rice) |
Coursera |
| Performance Engineering (MIT 6.172) |
MIT OCW |
MIT OCW |
| Parallel Computer Architecture (NPTEL) |
NPTEL (IIT Kanpur) |
NPTEL |
| Real-Time Systems (NPTEL) |
NPTEL (IIT Kharagpur) |
NPTEL |
| Concurrency is Not Parallelism (Rob Pike) |
YouTube |
YouTube |
| The Art of Multiprocessor Programming (talks) |
YouTube |
YouTube |
🔒 Concurrency Primitives
| Primitive |
What |
When |
| Mutex |
Mutual exclusion lock |
Protect shared state |
| Semaphore |
Counting lock |
Limit concurrent access |
| Condition Variable |
Wait for a condition |
Producer-consumer |
| Read-Write Lock |
Multiple readers OR one writer |
Read-heavy workloads |
| Spinlock |
Busy-wait lock |
Very short critical sections |
| Atomic Operations |
Lock-free single-variable ops |
Counters, flags |
| Barrier |
Wait for all threads |
Parallel phases |
| Channel |
Message passing (Go, Rust) |
CSP-style concurrency |
| Future/Promise |
Async result |
Non-blocking I/O |
🏗️ Concurrency Models
| Model |
Language/Runtime |
Key Idea |
| Threads + Locks |
C, C++, Java |
Shared memory, explicit synchronization |
| CSP (Communicating Sequential Processes) |
Go (goroutines + channels) |
Don't share memory; share by communicating |
| Actor Model |
Erlang, Akka, Elixir |
Isolated actors, message passing |
| Async/Await |
Python, JS, Rust, C# |
Cooperative multitasking, event loop |
| Data Parallelism |
CUDA, OpenMP, SIMD |
Same operation on many data points |
| Fork-Join |
Java ForkJoinPool, Rayon (Rust) |
Divide work, join results |
| Software Transactional Memory |
Haskell, Clojure |
Database-like transactions for memory |
📅 Scheduling
OS Scheduling Algorithms
| Algorithm |
Type |
Use Case |
| FIFO (First Come First Served) |
Non-preemptive |
Batch processing |
| Round Robin |
Preemptive, time-slice |
General-purpose OS |
| Priority Scheduling |
Preemptive |
Real-time systems |
| Shortest Job First (SJF) |
Non-preemptive |
Batch optimization |
| Multilevel Feedback Queue |
Preemptive, adaptive |
Linux CFS inspiration |
| Completely Fair Scheduler (CFS) |
Preemptive |
Linux default |
| Earliest Deadline First (EDF) |
Preemptive |
Hard real-time |
| Rate Monotonic (RM) |
Static priority |
RTOS, periodic tasks |
Real-Time Scheduling
→ See Embedded Track
| Concept |
What |
| Hard real-time |
Missing deadline = system failure |
| Soft real-time |
Missing deadline = degraded quality |
| Priority inversion |
Low-priority task blocks high-priority |
| Priority inheritance |
Temporarily boost blocking task |
| Watchdog timer |
Reset if task doesn't complete in time |
Resources
🔓 Lock-Free & Wait-Free Programming
| Resource |
Link |
| Herb Sutter: Lock-Free Programming (CppCon) |
YouTube |
| Jeff Preshing: Lock-Free Blog |
preshing.com |
| Memory Models (C++ atomics) |
YouTube |
| Fedor Pikus: C++ Atomics (CppCon) |
YouTube |
🐛 Common Concurrency Bugs
| Bug |
What |
Prevention |
| Race Condition |
Outcome depends on timing |
Locks, atomics, immutability |
| Deadlock |
Threads wait for each other forever |
Lock ordering, timeouts |
| Livelock |
Threads keep retrying, no progress |
Backoff strategies |
| Starvation |
Thread never gets CPU time |
Fair scheduling, priority |
| Priority Inversion |
High-priority blocked by low-priority |
Priority inheritance |
| ABA Problem |
Value changes A→B→A, CAS succeeds incorrectly |
Hazard pointers, epoch-based reclamation |
📖 Books
Cross-references: OS & Kernel · C++ · Rust · Go · Embedded Track