CSC364 Lecture 08

Introduction to Networked, Distributed, and Parallel Computing
Concurrency Patterns
(202602)

Transcript

1. Dr. Javier Gonzalez-Sanchez javiergs@calpoly.edu www.javiergs.info o ffi ce: 14 -227 CSC 364 Introduction to Networked, Distributed, and Parallel Computing Lecture 08. Concurrency Design Patterns 1
2. 2 De f inition A Concurrency Design P a ttern is a reus a ble solution to a common coordin a tion problem between multiple thre a ds. They describe: • How thre a ds inter a ct • How sh a red d a t a is a ccessed • How execution is coordin a ted s a fely P a tterns a re a bout structure a nd intent
3. 3 Why Concurrency Patterns? Concurrency mech a nisms (sem a phore, lock, synchronize, pool, a tomic) tell us how to synchronize. P a tterns expl a in when a nd why. With p a tterns: • Code is e a sier to re a son a bout • Beh a vior is predict a ble • Systems sc a le more s a fely
4. 4 Patterns 1. Producer–Consumer: Producers gener a te d a t a ; consumers t a ke d a t a using a sh a red bu ff er or queue. 2. Re a ders–Writers: Multiple re a ders c a n re a d simult a neously, but writers require exclusive a ccess.
5. Producer-Consumer Pattern 5
6. 6 Features • The Producer–Consumer p a ttern sep a r a tes the gener a tion of work from the processing of work. • One or more thre a ds produce t a sks. • One or more thre a ds consume t a sks. • They oper a te a t di ff erent speeds. • It provide decoupling between producers a nd consumers • S a fe sh a ring of a work bu ff er
7. 7 When to Use Producer–Consumer? Use when: • Work a rrives continuously • Processing is a synchronous • Producers a nd consumers should not block e a ch other unnecess a rily Ex a mples: • T a sk queues • Network mess a ge h a ndling
8. Objects Producer Consumer 1. 2. 3. 4. 5. Buffer
9. Concurrency Mechanisms Producer Consumer 1. 2. 3. 4. 5. Buffer 5 0
10. 1. The producer is gated by Semaphore A Producer Consumer 1. 2. 3. 4. 5. Buffer 5 0 ✅ 📦
11. 2. Producer acquires the buffer lock and produces Producer Consumer 1. 2. 3. 4. 5. Buffer 4 0 ✅ 📦
12. 3. Producer releases the lock and signals Semaphore B Producer Consumer 1. 2. 3. 4. 5. Buffer 4 1 📦
13. 4. The consumer is gated by Semaphore B Producer Consumer 1. 2. 3. 4. 5. Buffer 4 1 ✅ 📦
14. 5. Consumer acquires the buffer lock and consumes Producer Consumer 1. 2. 3. 4. 5. Buffer 4 0 ✅ 📦
15. 15 6. Consumer releases the lock and signals Semaphore A Producer Consumer 1. 2. 3. 4. 5. Buffer 5 0 📦
16. 16 Main
17. 17 Main
18. 18 Producer
19. 19 Producer
20. 20 Consumer
21. 21 Consumer
22. 22 Buffer (1/2)
23. 23 Buffer (2/2)
24. 24 Buffer (1/2)
25. 25 Buffer (1/2)
26. 26 In general Producer Consumer 1. 2. 3. 4. 5. Buffer 5 0 Producer Producer Consumer Consumer Consumer
27. Readers-Writers Pattern 27
28. 28 Features • Multiple thre a ds to re a d simult a neously • Re a ds do not block other re a ds. • Only one thre a d to write exclusively • Writes block everyone. • High perform a nce for re a d-he a vy worklo a ds • D a t a consistency during writes • Controlled a ccess b a sed on oper a tion type
29. 29 When to Use Readers–Writers? • Re a ds a re much more frequent th a n writes • Sh a red d a t a must rem a in consistent • Exclusive write a ccess is required Ex a mples: • Course C a t a log in a University System • Web server c a che • G a me st a tes • IDE - compiler
30. 30 Java ReadWriteLock A Re a dWriteLock is a lock with two di ff erent modes of a ccess: • Re a d lock → sh a red • Write lock → exclusive Why do we need it? • With a norm a l lock (synchronized or Reentr a ntLock) only one thre a d c a n a ccess the d a t a a t a time • Re a ders block other re a ders, even though they don’t modify a nything. Blocking re a ders from e a ch other is unnecess a ry a nd slow.
31. 31 Objects Reader 1. 2. 3. 4. 5. Buffer Writer 📄 📄 📄 📄 📄
32. 32 Concurrency Mechanisms Writer Reader 1. 2. 3. 4. 5. Buffer 📄 📄 📄 📄 📄
33. 33 1. Writer acquires the buffer lock and write Writer Reader 1. 2. 3. 4. 5. Buffer 📄 📜 📄 📄 📄 📄
34. 34 Concurrency Mechanisms Writer Reader 1. 2. 3. 4. 5. Buffer 📜 📄 📄 📄 📄
35. 35 3. Reader acquires the buffer lock and read Writer Reader 1. 2. 3. 4. 5. Buffer 📜 📄 📄 📄 📄
36. 36 Additional Readers are Allowed to Read Writer Reader 1. 2. 3. 4. 5. Buffer 📜 📄 📄 📄 📄
37. 37 Main
38. 38 Main
39. 39 Reader
40. 40 Reader
41. 41 Writer
42. 42 Writer
43. 43 CourseCatalog (1/2)
44. 44 CourseCatalog (1/2)
45. 45 CourseCatalog (2/2)
46. 46 CourseCatalog (2/2)
47. 47 In general Writer 1. 2. 3. 4. 5. Buffer 📜 📄 📄 📄 📄 Reader Reader Reader Reader Writer Writer Reader Reader Reader
48. Assignment 02 Team Ball Chase 48
49. 49 Concurrent Team Ball Chase • Two te a ms move on a grid: Te a m A a nd Te a m B. There is one sh a red b a ll, a nd only one pl a yer owns the b a ll a t a time. • The b a ll c a rrier tries to re a ch the opponent’s go a l. Other pl a yers move concurrently: te a mm a tes try to block ne a rby enemies or support the b a ll c a rrier, while enemies try to collide with the b a ll c a rrier to ste a l the b a ll.
50. 50 Path f inder Mode 1
51. 51 Path f inder 🧶 Mode 1
52. 52 Path f inder 🧶 Mode 1
53. 53 Path f inder 🧶 Mode 1
54. 54 Game Rules 1. Pl a yers m a y move one cell per turn in a ny of the 8 directions. 2. Other pl a yers a ct a s moving obst a cles. 3. No two pl a yers m a y occupy the s a me cell unless the move is a v a lid collision with the opposing b a ll c a rrier. 4. If a n enemy collides with the b a ll c a rrier, possession ch a nges. 5. When possession ch a nges, the te a m with the b a ll becomes the a tt a cking te a m. 6. If the b a ll c a rrier re a ches the opponent’s go a l, th a t te a m scores. 7. After a go a l, a ll pl a yers return to their origin a l st a rting positions, a nd the te a m th a t conceded the go a l st a rts with possession of the b a ll.
55. 55 Concurrent Behavior • Mode 1 — Turn-B a sed Concurrent Execution. E a ch pl a yer is upd a ted a s a t a sk once per turn. The g a me submits one t a sk per pl a yer, w a its for a ll t a sks to f inish, resolves moves, a nd then redr a ws the grid. • Mode 2 — Independent Pl a yer Thre a ds. E a ch pl a yer runs a s a n independent thre a d. Pl a yers m a y move a t di ff erent speeds, so some pl a yers m a y move more often th a n others. • Atomic / Lock / Sem a phore: protect b a ll ownership, pl a yer positions, a nd limited- a ccess a re a s. • Producer–Consumer: pl a yer t a sks produce movement events; the UI consumes them a nd redr a ws the grid. • Re a ders–Writers: m a ny pl a yers m a y re a d the grid st a te concurrently, but upd a tes to positions or possession require exclusive a ccess.
56. 56 Important Bugs to Watch For • If b a ll possession is not locked, two pl a yers m a y ste a l the b a ll a t the s a me time. • If position upd a tes a re not protected, two pl a yers m a y move into the s a me cell. • If movement locks a re a cquired incorrectly, pl a yers m a y block e a ch other forever.
57. 57 Console UI is f ine. Score A: 0 | B: 0 Ball: A1 Turn 12 G . . . . . . . G . A2 . . B3 . . . . . . A* . . B1 . . . A3 . . B2 . . G . . . . . . . G A* moves from (3,2) to (4,2) B2 moves from (5,3) to (4,2) <Messages> XX steals the ball or Team XX scores
58. 58 Questions
59. Lab
60. 60 Concurrent BFS Chase with Replanning . . . . . . . . X . # . X . . . . # . . . . # . T . # . . . . . . X . T = target (data) X = enemy (thread) # = wall (data) . = empty cell (data) Each enemy uses BFS to compute the shortest path to the target, moves one step along that path if the next cell is free, and then recomputes the path on its next turn. So, each enemy repeats this cycle independently: 1. Read the current grid 2. Run BFS from my position to T 3. Get the shortest path 4. Move only one step along that path 5. Sleep briefly 6. Repeat
61. CSC 364 Introduction to Introduction to Networked, Distributed, and Parallel Computing Javier Gonzalez-Sanchez, Ph.D. javiergs@calpoly.edu Winter 2026 Copyright. These slides can only be used as study material for the class CSC 364 at Cal Poly. They cannot be distributed or used for another purpose. 61