1. 1. Web Services 2. Concurrency and threads

2. Web History 1989:  Tim Berners-Lee (CERN) writes internal proposal to develop a distributed hypertext system.  Connects “a web of notes with links.”  Intended to help CERN physicists in large projects share and manage information 1990:  Tim BL writes a graphical browser for Next machines.

3. Web History (cont) 1992  NCSA server released  26 WWW servers worldwide 1993  Marc Andreessen releases first version of NCSA Mosaic browser  Mosaic version released for (Windows, Mac, Unix).  Web (port 80) traffic at 1% of NSFNET backbone traffic.  Over 200 WWW servers worldwide. 1994  Andreessen and colleagues leave NCSA to form “Mosaic Communications Corp” (predecessor to Netscape).

4. Internet Hosts

5. Web Servers Web server HTTP request HTTP response (content) Clients and servers communicate using the HyperText Transfer Protocol (HTTP)  Client and server establish TCP connection  Client requests content  Server responds with requested content  Client and server close connection (eventually) Current version is HTTP/1.1  RFC 2616, June, 1999. Web client (browser) http://www.w3.org/Protocols/rfc2616/rfc2616.html IP TCP HTTP Datagrams Streams Web content

6. Web Content Web servers return content to clients  content: a sequence of bytes with an associated MIME (Multipurpose Internet Mail Extensions) type Example MIME types  text/html HTML document  text/plain Unformatted text  application/postscript Postcript document  image/gif Binary image encoded in GIF format  image/jpeg Binary image encoded in JPEG format

7. Static and Dynamic Content The content returned in HTTP responses can be either static or dynamic.  Static content: content stored in files and retrieved in response to an HTTP request  Examples: HTML files, images, audio clips.  Request identifies content file  Dynamic content: content produced on-the-fly in response to an HTTP request  Example: content produced by a program executed by the server on behalf of the client.  Request identifies file containing executable code Bottom line: All Web content is associated with a file that is managed by the server.

8. URLs Each file managed by a server has a unique name called a URL (Universal Resource Locator) URLs for static content:  http://reed.cs.depaul.edu:80/index.html  http://reed.cs.depaul.edu/index.html  http://reed.cs.depaul.edu  Identifies a file called index.html, managed by a Web server at reed.cs.depaul.edu that is listening on port 80. URLs for dynamic content:  http://riely373.cdm.depaul.edu:8000/cgi-bin/adder?15000&213  Identifies an executable file called adder, managed by a Web server at riely373.cdm.depaul.edu that is listening on port 8000, that should be called with two argument strings: 15000 and 213.

9. How Clients and Servers Use URLs Example URL: http://www.depaul.edu:80/index.html Clients use prefix ( http://www.depaul.edu:80 ) to infer:  What kind of server to contact (Web server)  Where the server is ( www.depaul.edu )  What port it is listening on (80) Servers use suffix ( /index.html ) to:  Determine if request is for static or dynamic content.  No hard and fast rules for this.  Convention: executables reside in cgi-bin directory  Find file on file system.  Initial “ / ” in suffix denotes home directory for requested content.  Minimal suffix is “ / ”, which all servers expand to some default home page (e.g., index.html ).

10. Anatomy of an HTTP Transaction $ telnet reed.cs.depaul.edu 80 Trying 140.192.39.42... Connected to reed.cti.depaul.edu. Escape character is '^]'. GET / HTTP/1.1 host: reed.cs.depaul.edu HTTP/1.1 200 OK Server: Apache-Coyote/1.1 Accept-Ranges: bytes ETag: W/"2285-1357855910000" Last-Modified: Thu, 10 Jan 2013 22:11:50 GMT Content-Type: text/html Content-Length: 2285 Date: Mon, 04 Mar 2013 04:01:00 GMT...

11. HTTP Requests HTTP request is a request line, followed by zero or more request headers Request line:  is HTTP version of request ( HTTP/1.0 or HTTP/1.1 )  is typically URL for proxies, URL suffix for servers.  A URL is a type of URI (Uniform Resource Identifier)  See http://www.ietf.org/rfc/rfc2396.txt  is either GET, POST, OPTIONS, HEAD, PUT, DELETE, or TRACE.

12. HTTP Requests (cont) HTTP methods:  GET : Retrieve static or dynamic content  Arguments for dynamic content are in URI  Workhorse method (99% of requests)  POST : Retrieve dynamic content  Arguments for dynamic content are in the request body  OPTIONS : Get server or file attributes  HEAD : Like GET but no data in response body  PUT : Write a file to the server!  DELETE : Delete a file on the server!  TRACE : Echo request in response body  Useful for debugging. Request headers: :  Provide additional information to the server.

13. HTTP Versions Major differences between HTTP/1.1 and HTTP/1.0  HTTP/1.0 uses a new connection for each transaction.  HTTP/1.1 also supports persistent connections  multiple transactions over the same connection  Connection: Keep-Alive  HTTP/1.1 requires HOST header  Host: www.depaul.eduwww.depaul.edu  Makes it possible to host multiple websites at single Internet host  HTTP/1.1 supports chunked encoding (described later)  Transfer-Encoding: chunked  HTTP/1.1 adds additional support for caching

14. HTTP Responses HTTP response is a response line followed by zero or more response headers. Response line:  is HTTP version of the response.  is numeric status.  is corresponding English text.  200 OKRequest was handled without error  301MovedProvide alternate URL  403ForbiddenServer lacks permission to access file  404Not foundServer couldn’t find the file. Response headers: :  Provide additional information about response  Content-Type: MIME type of content in response body.  Content-Length: Length of content in response body.

15. GET Request From Chrome Browser GET / HTTP/1.1\r\n Host: reed.cs.depaul.edu\r\n Connection: keep-alive\r\n Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8\r\n User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.22 (KHTML, like Gecko) Chrome/25.0.1364.97 Safari/537.22\r\n Accept-Encoding: gzip,deflate,sdch\r\n Accept-Language: en-US,en;q=0.8\r\n Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.3\r\n Cookie:__utma=114012434.756988690.1360702406.1360702406.1360874291.2; __utmz=114012434.1360874291.2.2.utmcsr=cdm.depaul.edu|utmccn=(referral )|utmcmd=referral|utmcct=/academics/Pages/bs%20computerscience%20stand ard.aspx\r\n \r\n URI is just the suffix, not the entire URL

16. GET Response From Apache Server HTTP/1.1 200 OK Server: Apache-Coyote/1.1\r\n Accept-Ranges: bytes\r\n ETag: W/”2285-1357855910000”\r\n Last-Modified: Thu, 10 Jan 2013 22:11:50 GMT\r\n Content-Type: test/html\r\n Content-Length: 2285\r\n Date: Mon, 04 Mar 2013 04:58:40 GMT\r\n \r\n \n...

17. Tiny Web Server Tiny Web server described in text  Tiny is a sequential Web server.  Serves static and dynamic content to real browsers.  text files, HTML files, GIF and JPEG images.  226 lines of commented C code.  Not as complete or robust as a real web server

18. Tiny Operation Read request from client Split into method / uri / version  If not GET, then return error If URI contains “ cgi-bin ” then serve dynamic content  Fork process to execute program Otherwise serve static content  Copy file to output

19. Tiny Serving Static Content  Serve file specified by filename  Use file metadata to compose header  “Read” file via mmap  Write to output /* Send response headers to client */ get_filetype(filename, filetype); sprintf(buf, "HTTP/1.0 200 OK\r\n"); sprintf(buf, "%sServer: Tiny Web Server\r\n", buf); sprintf(buf, "%sContent-length: %d\r\n", buf, filesize); sprintf(buf, "%sContent-type: %s\r\n\r\n", buf, filetype); Rio_writen(fd, buf, strlen(buf)); /* Send response body to client */ srcfd = Open(filename, O_RDONLY, 0); srcp = Mmap(0, filesize, PROT_READ, MAP_PRIVATE, srcfd, 0); Close(srcfd); Rio_writen(fd, srcp, filesize); Munmap(srcp, filesize); From tiny.c

20. Serving Dynamic Content ClientServer Client sends request to server. If request URI contains the string “ /cgi-bin ”, then the server assumes that the request is for dynamic content. GET /cgi-bin/env.pl HTTP/1.1

21. Serving Dynamic Content (cont) ClientServer The server creates a child process and runs the program identified by the URI in that process env.pl fork/exec

22. Serving Dynamic Content (cont) ClientServer The child runs and generates the dynamic content. The server captures the content of the child and forwards it without modification to the client env.pl Content

23. Issues in Serving Dynamic Content How does the client pass program arguments to the server? How does the server pass these arguments to the child? How does the server pass other info relevant to the request to the child? How does the server capture the content produced by the child? These issues are addressed by the Common Gateway Interface (CGI) specification. ClientServer Content Request Create env.pl

24. CGI Because the children are written according to the CGI spec, they are often called CGI programs. Because many CGI programs are written in Perl, they are often called CGI scripts. However, CGI really defines a simple standard for transferring information between the client (browser), the server, and the child process.

25. The cdmlinux addition portal input URL Output page hostportCGI program args

26. Serving Dynamic Content With GET Question: How does the client pass arguments to the server? Answer: The arguments are appended to the URI Can be encoded directly in a URL typed to a browser or a URL in an HTML link  http://cdmlinux.cdm.depaul.edu/cgi-bin/adder?n1=4&n2=7  adder is the CGI program on the server that will do the addition.  argument list starts with “?”  arguments separated by “&”  spaces represented by “+” or “%20” URI often generated by an HTML form X Y

27. Serving Dynamic Content With GET URL:  cgi-bin/adder?4&7 Result displayed on browser: Welcome to THE Internet addition portal. The answer is: 4+7=11 Thanks for visiting!

28. Serving Dynamic Content With GET Question: How does the server pass these arguments to the child? Answer: In environment variable QUERY_STRING  A single string containing everything after the “?”  For add: QUERY_STRING = “ 4&7 ”

29. Additional CGI Environment Variables General  SERVER_SOFTWARE  SERVER_NAME  GATEWAY_INTERFACE (CGI version) Request-specific  SERVER_PORT  REQUEST_METHOD ( GET, POST, etc)  QUERY_STRING (contains GET args)  REMOTE_HOST (domain name of client)  REMOTE_ADDR (IP address of client)  CONTENT_TYPE (for POST, type of data in message body, e.g., text/html )  CONTENT_LENGTH (length in bytes)

30. Even More CGI Environment Variables In addition, the value of each header of type type received from the client is placed in environment variable HTTP_ type  Examples (any “-” is changed to “_”) :  HTTP_ACCEPT  HTTP_HOST  HTTP_USER_AGENT

31. Serving Dynamic Content With GET Question: How does the server capture the content produced by the child? Answer: The child generates its output on stdout. Server uses dup2 to redirect stdout to its connected socket.  Notice that only the child knows the type and size of the content. Thus the child (not the server) must generate the corresponding headers. /* Make the response body */ sprintf(content, "Welcome to add.com: "); sprintf(content, "%sTHE Internet addition portal.\r\n ", content); sprintf(content, "%sThe answer is: %s\r\n ", content, msg); sprintf(content, "%sThanks for visiting!\r\n", content); /* Generate the HTTP response */ printf("Content-length: %u\r\n", (unsigned) strlen(content)); printf("Content-type: text/html\r\n\r\n"); printf("%s", content); From adder.c

32. Serving Dynamic Content With GET HTTP request sent by client HTTP response generated by the server HTTP response generated by the CGI program $ telnet riely373.cdm.depaul.edu 8000 Trying 140.192.39.11... Connected to riely373.cdm.depaul.edu. Escape character is '^]'. GET /cgi-bin/adder?4&7 HTTP/1.0 HTTP/1.0 200 OK Server: Tiny Web Server Content-length: 97 Content-type: text/html Welcome to THE Internet addition portal. The answer is: 4 + 7 = 11 Thanks for visiting! Connection closed by foreign host. $

33. Tiny Serving Dynamic Content  Fork child to execute CGI program  Change stdout to be connection to client  Execute CGI program with execve /* Return first part of HTTP response */ sprintf(buf, "HTTP/1.0 200 OK\r\n"); Rio_writen(fd, buf, strlen(buf)); sprintf(buf, "Server: Tiny Web Server\r\n"); Rio_writen(fd, buf, strlen(buf)); if (Fork() == 0) { /* child */ /* Real server would set all CGI vars here */ setenv("QUERY_STRING", cgiargs, 1); Dup2(fd, STDOUT_FILENO); /* Redirect stdout to client */ Execve(filename, emptylist, environ);/* Run CGI prog */ } Wait(NULL); /* Parent waits for and reaps child */ From tiny.c

34. Proxies A proxy is an intermediary between a client and an origin server.  To the client, the proxy acts like a server.  To the server, the proxy acts like a client. ClientProxy Origin Server 1. Client request 2. Proxy request 3. Server response4. Proxy response

35. Why Proxies? Can perform useful functions as requests and responses pass by  Examples: Caching, logging, anonymization, filtering, transcoding Client A Proxy cache Origin Server Request foo.html foo.html Client B Request foo.html foo.html Fast inexpensive local network Slower more expensive global network

36. For More Information Study the Tiny Web server described in your text  Tiny is a sequential Web server.  Serves static and dynamic content to real browsers.  text files, HTML files, GIF and JPEG images.  220 lines of commented C code.  Also comes with an implementation of the CGI script for the add.com addition portal. See the HTTP/1.1 standard:  http://www.w3.org/Protocols/rfc2616/rfc2616.html

37. 2. Concurrency and threads

38. Client / Server Session Iterative Echo Server ClientServer socket bind listen rio_readlineb rio_writenrio_readlineb rio_writen Connection request rio_readlineb close EOF Await connection request from next client open_listenfd open_clientfd acceptconnect

39. Iterative Servers Iterative servers process one request at a time client 1serverclient 2 connect accept connect write read call read close accept write read close Wait for Client 1 call read write ret read write ret read

40. Creating Concurrent Flows  Allow server to handle multiple clients simultaneously 1. Processes  Kernel automatically interleaves multiple logical flows  Each flow has its own private address space 2. Threads  Kernel automatically interleaves multiple logical flows  Each flow shares the same address space 3. I/O multiplexing with select()  Programmer manually interleaves multiple logical flows  All flows share the same address space  Relies on lower-level system abstractions

41. Concurrent Servers: Multiple Processes Spawn separate process for each client client 1serverclient 2 call connect call accept call read ret connect ret accept call connect call fgets fork child 1 User goes out to lunch Client 1 blocks waiting for user to type in data call accept ret connect ret accept call fgets writefork call read child 2 write call read end read close...

42. Review: Iterative Echo Server int main(int argc, char **argv) { int listenfd, connfd; int port = atoi(argv[1]); struct sockaddr_in clientaddr; int clientlen = sizeof(clientaddr); listenfd = Open_listenfd(port); while (1) { connfd = Accept(listenfd, (SA *)&clientaddr, &clientlen); echo(connfd); Close(connfd); } exit(0); }  Accept a connection request  Handle echo requests until client terminates

43. int main(int argc, char **argv) { int listenfd, connfd; int port = atoi(argv[1]); struct sockaddr_in clientaddr; int clientlen=sizeof(clientaddr); Signal(SIGCHLD, sigchld_handler); listenfd = Open_listenfd(port); while (1) { connfd = Accept(listenfd, (SA *) &clientaddr, &clientlen); if (Fork() == 0) { Close(listenfd); /* Child closes its listening socket */ echo(connfd); /* Child services client */ Close(connfd); /* Child closes connection with client */ exit(0); /* Child exits */ } Close(connfd); /* Parent closes connected socket (important!) */ } Process-Based Concurrent Server Fork separate process for each client Does not allow any communication between different client handlers

44. Process-Based Concurrent Server (cont) void sigchld_handler(int sig) { while (waitpid(-1, 0, WNOHANG) > 0) ; return; }  Reap all zombie children

45. Process Execution Model  Each client handled by independent process  No shared state between them  Both parent & child have copies of listenfd and connfd  Parent must close connfd  Child must close listenfd Client 1 Server Process Client 2 Server Process Listening Server Process Connection Requests Client 1 dataClient 2 data

46. Concurrent Server: accept Illustrated listenfd(3) Client 1. Server blocks in accept, waiting for connection request on listening descriptor listenfd clientfd Server listenfd(3) Client clientfd Server 2. Client makes connection request by calling and blocking in connect Connection request listenfd(3) Client clientfd Server 3. Server returns connfd from accept. Forks child to handle client. Client returns from connect. Connection is now established between clientfd and connfd Server Child connfd(4)

47. Implementation Must-dos With Process-Based Designs Listening server process must reap zombie children  to avoid fatal memory leak Listening server process must close its copy of connfd  Kernel keeps reference for each socket/open file  After fork, refcnt(connfd) = 2  Connection will not be closed until refcnt(connfd) == 0

48. Pros and Cons of Process-Based Designs + Handle multiple connections concurrently + Clean sharing model  descriptors (no)  file tables (yes)  global variables (no) + Simple and straightforward – Additional overhead for process control – Nontrivial to share data between processes  Requires IPC (interprocess communication) mechanisms  FIFO’s (named pipes), System V shared memory and semaphores

49. Approach #2: Multiple Threads Very similar to approach #1 (multiple processes)  but, with threads instead of processes

50. Traditional View of a Process Process = process context + code, data, and stack shared libraries run-time heap 0 read/write data Program context: Data registers Condition codes Stack pointer (SP) Program counter (PC) Kernel context: VM structures Descriptor table brk pointer Code, data, and stack read-only code/data stack SP PC brk Process context

51. Alternate View of a Process Process = thread + code, data, and kernel context shared libraries run-time heap 0 read/write data Thread context: Data registers Condition codes Stack pointer (SP) Program counter (PC) Code and Data read-only code/data stack SP PC brk Thread (main thread) Kernel context: VM structures Descriptor table brk pointer

52. A Process With Multiple Threads Multiple threads can be associated with a process  Each thread has its own logical control flow  Each thread shares the same code, data, and kernel context  Share common virtual address space (inc. stacks)  Each thread has its own thread id (TID) shared libraries run-time heap 0 read/write data Thread 1 context: Data registers Condition codes SP1 PC1 Shared code and data read-only code/data stack 1 Thread 1 (main thread) Kernel context: VM structures Descriptor table brk pointer Thread 2 context: Data registers Condition codes SP2 PC2 stack 2 Thread 2 (peer thread)

53. Logical View of Threads Threads associated with process form a pool of peers  Unlike processes which form a tree hierarchy P0 P1 sh foo bar T1 Process hierarchy Threads associated with process foo T2 T4 T5 T3 shared code, data and kernel context

54. Thread Execution Single Core Processor  Simulate concurrency by time slicing Multi-Core Processor  Can have true concurrency Time Thread AThread BThread C Thread AThread BThread C Run 3 threads on 2 cores

55. Threads vs. Processes How threads and processes are similar  Each has its own logical control flow  Each can run concurrently with others (possibly on different cores)  Each is context switched How threads and processes are different  Threads share code and some data  Processes (typically) do not  Threads are somewhat less expensive than processes  Process control (creating and reaping) is twice as expensive as thread control  Linux numbers: –~20K cycles to create and reap a process –~10K cycles (or less) to create and reap a thread

56. Posix Threads (Pthreads) Interface Pthreads: Standard interface for ~60 functions that manipulate threads from C programs  Creating and reaping threads  pthread_create()  pthread_join()  Determining your thread ID  pthread_self()  Terminating threads  pthread_cancel()  pthread_exit()  exit() [terminates all threads], RET [terminates current thread]  Synchronizing access to shared variables  pthread_mutex_init  pthread_mutex_[un]lock  pthread_cond_init  pthread_cond_[timed]wait

57. /* thread routine */ void *thread(void *vargp) { printf("Hello, world!\n"); return NULL; } The Pthreads "hello, world" Program /* * hello.c - Pthreads "hello, world" program */ #include "csapp.h" void *thread(void *vargp); int main() { pthread_t tid; Pthread_create(&tid, NULL, thread, NULL); Pthread_join(tid, NULL); exit(0); } Thread attributes (usually NULL) Thread arguments (void *p) return value (void **p)

58. Execution of Threaded“hello, world” main thread peer thread return NULL; main thread waits for peer thread to terminate exit() terminates main thread and any peer threads call Pthread_create() call Pthread_join() Pthread_join() returns printf() (peer thread terminates) Pthread_create() returns

59. Thread-Based Concurrent Echo Server int main(int argc, char **argv) { int port = atoi(argv[1]); struct sockaddr_in clientaddr; int clientlen=sizeof(clientaddr); pthread_t tid; int listenfd = Open_listenfd(port); while (1) { int *connfdp = Malloc(sizeof(int)); *connfdp = Accept(listenfd, (SA *) &clientaddr, &clientlen); Pthread_create(&tid, NULL, echo_thread, connfdp); }  Spawn new thread for each client  Pass it copy of connection file descriptor  Note use of Malloc()!  Without corresponding Free()

60. Thread-Based Concurrent Server (cont) /* thread routine */ void *echo_thread(void *vargp) { int connfd = *((int *)vargp); Pthread_detach(pthread_self()); Free(vargp); echo(connfd); Close(connfd); return NULL; }  Run thread in “detached” mode  Runs independently of other threads  Reaped when it terminates  Free storage allocated to hold clientfd  “Producer-Consumer” model

61. Threaded Execution Model  Multiple threads within single process  Some state between them  File descriptors Client 1 Server Client 2 Server Listening Server Connection Requests Client 1 dataClient 2 data

62. Potential Form of Unintended Sharing main thread peer 1 while (1) { int connfd = Accept(listenfd, (SA *) &clientaddr, &clientlen); Pthread_create(&tid, NULL, echo_thread, (void *) &connfd); } connfd Main thread stack vargp Peer 1 stack vargp Peer 2 stack peer 2 connfd = connfd 1 connfd = *vargp connfd = connfd 2 connfd = *vargp Race! Why would both copies of vargp point to same location?

63. Could this race occur? int i; for (i = 0; i < 100; i++) { Pthread_create(&tid, NULL, thread, &i); } Race Test  If no race, then each thread would get different value of i  Set of saved values would consist of one copy each of 0 through 99. Main void *thread(void *vargp) { int i = *((int *)vargp); Pthread_detach(pthread_self()); save_value(i); return NULL; } Thread

64. Experimental Results The race can really happen! No Race Multicore server Single core laptop

65. Issues With Thread-Based Servers Must run “detached” to avoid memory leak.  At any point in time, a thread is either joinable or detached.  Joinable thread can be reaped and killed by other threads.  must be reaped (with pthread_join ) to free memory resources.  Detached thread cannot be reaped or killed by other threads.  resources are automatically reaped on termination.  Default state is joinable.  use pthread_detach(pthread_self()) to make detached. Must be careful to avoid unintended sharing.  For example, passing pointer to main thread’s stack Pthread_create(&tid, NULL, thread, (void *)&connfd); All functions called by a thread must be thread-safe  Stay tuned

66. Pros and Cons of Thread-Based Designs + Easy to share data structures between threads  e.g., logging information, file cache. + Threads are more efficient than processes. – Unintentional sharing can introduce subtle and hard-to- reproduce errors!  The ease with which data can be shared is both the greatest strength and the greatest weakness of threads.  Hard to know which data shared & which private  Hard to detect by testing  Probability of bad race outcome very low  But nonzero!  Future lectures

67. Event-Based Concurrent Servers Using I/O Multiplexing Use library functions to construct scheduler within single process Server maintains set of active connections  Array of connfd’s Repeat:  Determine which connections have pending inputs  If listenfd has input, then accept connection  Add new connfd to array  Service all connfd’s with pending inputs Details in book

68. I/O Multiplexed Event Processing 10 clientfd 7 4 12 5 0 1 2 3 4 5 6 7 8 9 Active Inactive Active Never Used listenfd = 3 10 clientfd 7 4 12 5 listenfd = 3 Active Descriptors Pending Inputs Read

69. Pros and Cons of I/O Multiplexing + One logical control flow. + Can single-step with a debugger. + No process or thread control overhead.  Design of choice for high-performance Web servers and search engines. – Significantly more complex to code than process- or thread- based designs. – Hard to provide fine-grained concurrency  E.g., our example will hang up with partial lines. – Cannot take advantage of multi-core  Single thread of control

70. Approaches to Concurrency Processes  Hard to share resources: Easy to avoid unintended sharing  High overhead in adding/removing clients Threads  Easy to share resources: Perhaps too easy  Medium overhead  Not much control over scheduling policies  Difficult to debug  Event orderings not repeatable I/O Multiplexing  Tedious and low level  Total control over scheduling  Very low overhead  Cannot create as fine grained a level of concurrency  Does not make use of multi-core