Communication

1. CSCE455/855Distributed Operating Systems Communication Dr. Ying Lu ylu@cse.unl.edu

2. Giving credit where credit is due: CSCE455/855Distributed Operating Systems • Most of the lecture notes are from the textbook companion website • Some of the lecture notes are based on slides created by Dr. Krzyzanowski at Rutgers University • I have modified them and added new slides

3. Layered Protocols (I) Reference model for networked communication

4. Layered Protocols (2)

5. Client-Server Communication • Assume that you are developing a client-server application: • How to let the two processes (client and server) located on two machines communicate with each other? • Socket programming: using functions like connect(sd, (struct sockaddr *)&sin, sizeof(sin)), write(sd, buf, strlen(buf)) etc.

6. Remote Procedure Calls (RPC) • Avoid explicit message exchange between processes • Basic idea is to allow a process on a machine to call procedures on a remote machine • Make a remote procedure possibly look like a local one • Original paper on RPC: • A. Birrell, B Nelson, “Implementing Remote Procedure Calls”, ACM Symposium on Operating System Principles, 1984

7. Conventional Procedure Call • How are parameters passed in a local procedure call • E.g., #include <sys/types.h> #include <unistd.h> ... char buf[20]; size_tnbytes; ssize_tbytes_read; intfd; ... nbytes = sizeof(buf); bytes_read = read(fd, buf, nbytes); ...

8. Conventional Procedure Call Figure 4-5. (a) Parameter passing in a local procedure call: the stack before the call to read. (b) The stack while the called procedure is active.

9. Remote Procedure Calls (RPC) • How are parameter passed in a remote procedure call, while making it look like a local procedure call?

10. Client and Server Stubs Principle of RPC between a client and server program.

11. Steps of a Remote Procedure Call • Client procedure calls client stub in normal way • Client stub builds message, calls local OS • Client's OS sends message to remote OS • Remote OS gives message to server stub • Server stub unpacks parameters, calls server • Server does work, returns result to the stub • Server stub packs it in message, calls local OS • Server's OS sends message to client's OS • Client's OS gives message to client stub • Stub unpacks result, returns to client

12. Passing Value Parameters (1) 2-8 Steps involved in doing remote computation through RPC

13. Passing Value Parameters (2)

14. Passing Value Parameters (3) • Original message on the Pentium (little-endian) • The message after receipt on the SPARC (big-endian) Note: the little numbers in boxes indicate the address of each byte

15. Passing Value Parameters (3) • Original message on the Pentium (little-endian) • The message after receipt on the SPARC (big-endian) • The message after being inverted (integer 5, string: “LLIJ”) Note: the little numbers in boxes indicate the address of each byte

16. Passing reference parameters Machine B Machine A Copy value a and contents of loc b into a’ and loc b’ a a’ b b’ Return Copy contents of loc b’ into b Call foo(a, &b’ ) foo(a, &b ) • What is Call By Value and Call By Refernce? • Example: call foo(int, int * ) or read(fd, buf, nbytes) • Call by copy/restore • The dreaded “pointer problem” • Linked list • Complex graph

17. Marshalling Marshalling transferring data structure used in remote procedure call from one address space to another. Define a “network format”, for example following XDR (eXternal Data Representation) standard http://www.ietf.org/rfc/rfc1832.txt Values must cross the network Machine formats differ • Integer byte order • Little-endian or big-endian • Floating point format • IEEE 754 or not

18. RPC: The basic mechanism • Client calls a local procedure on the client stub • The client stub acts as a proxy and marshalls the call and the args. • The client stub send this to the remote system (via TCP/UDP) • The server stub unmarshalls the call and args from the client • The server stub calls the actual procedure on the server • The server stub marshalls the reply and sends it back to the client Server process Client process Client routines Server routines 1 5 Client stub Server stub 2 4 RPC runtime RPC runtime Process Process kernel kernel 3 6 Network routines Network routines Source: R. Stevens, Unix Network Programming (IPC) Vol 2, 1998

19. Example1: A Time Server Interface struct time { int seconds; int minutes; int hours; int day; int month; int year; char timezone[4]; } int gettime(t); struct time *t; int settime(t); struct time *t;

20. Example1: Client Stub for Settime int settime(t); struct time *t; { char *p, message[32]; int stat; p = message; p = put_int(p, SETTIME); p = put_int(p, t->seconds); p = put_int(p, t->minutes); p = put_int(p, t->hours); p = put_int(p, t->day); p = put_int(p, t->month); p = put_int(p, t->year); p = put_string(p, t->timezone, 4); stat = do_operation(“time_server”, message, 32); if(stat == SUCCESS) get_int(message, &stat); return(stat); }

21. Example1: Server Stub (1) • p = get_int(p, &t.minutes); • p = get_int(p, &t.hours); • p = get_int(p, &t.day); • p = get_int(p, &t.month); • p = get_int(p, &t.year); • p = get_string(p, &t.timezone, 4); • len = settime(&t); • put_int(message, len); • len = 4; • break; case GETTIME: /* code for unmarshalling and calling gettime */ } send_reply(message, len); } } void main_loop() { char *p, message[32]; int len, op_code; struct time t; for(;;) { len = receive_request(message, 32); if(len < 4) { /* error handling code */ } p = message; p = get_int(p, op_code); switch(op_code) { case SETTIME: if (len < 32) { /* error handling code */ } p = get_int(p, &t.seconds);

22. Writing a Client and a Server Figure 4-12. The steps in writing a client and a server in DCE RPC DCE: Distributed Computing Environment

23. Binding a Client to a Server (1) • Registration of a server makes it possible for a client to locate the server and bind to it. • Server location is done in two steps: • Locate the server’s machine. • Locate the server on that machine.

24. Binding a Client to a Server (2) Figure 4-13. Client-to-server binding in DCE.

25. Asynchronous RPC (1) 2-12 • The interconnection between client and server in a traditional RPC • The interaction using asynchronous RPC

26. Asynchronous RPC (2) 2-13 A client and server interacting through two asynchronous RPCs

27. LPC v.s. RPC • Global variables • Client and server fail independently • RPC: requires code to deal with server crashes

28. When Things Go Wrong • Semantics of remote procedure calls • Local procedure call: exactly once • How many times a remote procedure call may be called? • A remote procedure call may be called: • 0 time: server crashed or server process died before executing server code • 1 time: everything worked well • 1 or more: due to excess latency or lost reply from server, client retransmitted • Exactly once may be difficult to achieve with RPC

29. RPC Semantics • Most RPC systems will offer either: • at least once semantics • or at most once semantics • Understand application: • Illustrate some applications that “at least once” is suitable? • Idempotent functions: may be run any number of times without harm • Illustrate some applications that “at most once” is suitable?

30. Useful Links for RPC • RFC 1831: RPC Specification • RFC 1832: XDR Specification

31. In-Class Exercises (I) • C has a construction called a union, in which a field of a record (called a struct in C) can hold any one of several alternatives. At run time, there is no sure-fire way to tell which one is in there. Does this feature of C have any implications for remote procedure call? Explain your answer. • If the runtime system cannot tell what type value is in the field, it cannot marshal it correctly. Thus unions cannot be tolerated in an RPC system unless there is a tag field that unambiguously tells what the variant field holds.

32. In-Class Exercises (II) 2. One way to handle parameter conversion in RPC systems is to have each machine send parameters in its native representation, with the other one doing the translation, if need be. The native system could be indicated by a code in the first byte. However, since locating the first byte in the first word is precisely the problem, can this actually work? First of all, when one computer sends byte 0, it always arrives in byte 0.Thus the destination computer can simply access byte 0 (using a byte instruction) and the code will be in it. An alternative scheme is to put the code in all the bytes of the first word. Then no matter which byte is examined, the code will be there.

34. Appendix

35. Example: Writing a client and a server (1) /* interface.x */ /* Example Interface Definition */ struct square_in { long arg; }; struct square_out { long result; }; program SQUARE_PROG { version SQUARE_VERS { square_out SQUAREPROC( square_in ) = 1; /* Procedure number = 1 */ } = 1; /* Version number = 1 */ } = 0x31230000; /* program number */ Source: R. Stevens, Unix Network Programming (IPC) Vol 2, 1998

36. Example: Writing a client and a server (2) interface.x rpcgen interface.h Client Main interface_clnt.c (client stub) interface_xdr.c interface_svc.c (Server stub) Server.c Runtime lib Client Server Source: R. Stevens, Unix Network Programming (IPC) Vol 2, 1998

37. Unix Network Programming UNIX Network Programming, Volume 2, Second Edition: Interprocess Communications, Prentice Hall, 1999, ISBN 0-13-081081-9