ØMQ was created to address the need for cheaper and more efficient connectivity between distributed systems as hardware capabilities increased. It is an intelligent socket library that provides various connection patterns for messaging between processes. ØMQ aims to be fast, small, multiplatform, and multilingual. It uses common patterns like request-reply and publish-subscribe to allow processes to communicate simply and efficiently over in-process, inter-process, and network connections. The ØMQ community continues to enhance and support the open source library.

1. Introduction to ZeroMQ 1 December 2011 Pieter Hintjens, iMatix

4. Broker-based messaging

6. It had to be really fast and really simple

8. Many kinds of connection patterns

9. Multiplatform, multi-language (30+)

10. Fast (8M msg/sec, 30usec latency)

11. Small (20K lines of C++ code)

12. Open source LGPL (large community)

13. ØMQ Hello World import org.zeromq.ZMQ; public class hwclient { public static void main (String[] args){ ZMQ.Context context = ZMQ.context (1); ZMQ.Socket socket = context.socket (ZMQ.REQ); socket.connect ("tcp://localhost:5555"); socket.send ("Hello", 0); System.out.println (socket.recv(0)); } } import org.zeromq.ZMQ; public class hwserver { public static void main (String[] args) { ZMQ.Context context = ZMQ.context(1); ZMQ.Socket socket = context.socket(ZMQ.REP); socket.bind ("tcp://*:5555"); while (true) { byte [] request = socket.recv (0); socket.send("World", 0); } } }

14. Request-Reply Pattern

15. Publish-Subscribe Pattern

16. Pipeline Pattern

17. Simple ØMQ Application

19. Processes on one box (ipc://)

20. Processes on one network (tcp://)

21. Multicast group (pgm://)

22. Multihop ØMQ Application

23. Typical ØMQ Design

25. Multicast (PUB)

26. Fair-queuing (REP, SUB, PULL, DEALER)

27. Explicit addressing (ROUTER)

28. Unicast (PAIR)

30. One socket connect to many sockets

31. Queuing at sender and receiver

32. Automatic TCP (re)connect

33. Zero-copy for large messages

35. Design app as message-driven tasks

36. Each task reads from 1..n sockets

37. Tasks can talk over inproc://

38. Tasks can be split into processes over tcp://

39. No wait states, no locks, full CPU use

40. Scalable to any number of cores

42. Move to faster language where needed (C)

43. Run on arbitrary platforms (Windows, Android)

44. Scale to arbitrary sizes (2 cores, 16 cores...)

45. No per-core or per-seat licensing

46. Easy to experiment and learn

48. Small protocols for main flows

49. ØMQ patterns for main flows

50. Break components up for performance

51. Profile and test

52. Improve incrementally over many cycles

54. 2004 – AMQP standard for JPMorganChase

55. 2005 – OpenAMQ message broker/client

56. 2008 – ØMQ/0.x for the avant-garde

57. 2009 – ØMQ/1.x for pioneers (finance)

58. 2010 – ØMQ/2.x for early adopters (foss)

59. 2011 – ØMQ/3.x for mass market (cloud)

61. 1,000 people on dev list, 120 on IRC

63. Language bindings (35 or more)

66. zero.mq – community wiki

67. #zeromq – IRC channel on Freenode

68. zeromq-dev – email list on zeromq.org

69. github.com/zeromq – git repositories

70. zguide.zeromq.org – user guide

71. api.zeromq.org – reference manual