Scalability Issues: Cure now, prevent later

Slides for my Future Of Web Apps talk about scalability in Prague.

Transcript

1. View Slide

2. Hi  
   my  name  
   is  Thijs
   

   View Slide

3. THIJS
   

   View Slide

4. View Slide

5. TAZE
   

   View Slide

6. Excited  
   to  be  
   here!
   

   View Slide

7. I’m  
   an  evangelist  at
   

   View Slide

8. I’m  
   a  board  member  
   at
   

   View Slide

9. View Slide

10. View Slide

11. It  works  on  my  computer
    

    View Slide

12. View Slide

13. It’s  the  internet’s  fault
    

    View Slide

14. Developer  !=  web  developer
    

    View Slide

15. HTTP?  HUH?
    

    View Slide

16. View Slide

17. View Slide

18. View Slide

19. What  do  you  do  when  things  go  wrong?
    

    View Slide

20. Do  a  complete  rewrite  !
    

    View Slide

21. Absolutely  
    nothing!
    

    View Slide

22. Throw  servers  
    at  the  problem
    

    View Slide

23. Bigger  box
    

    View Slide

24. More  boxes
    

    View Slide

25. Reduce  
    impact
    

    View Slide

26. What  kind  of  infrastructure  do  you  need?
    

    View Slide

27. One  big  
    box?
    

    View Slide

28. Scale  across  
    100  nodes?
    

    View Slide

29. That  big  F5  
    loadbalancer?
    

    View Slide

30. Only  the  DB  
    gets  
    hammered
    

    View Slide

31. Lots  of  staCc  
    traffic
    

    View Slide

32. Shared  hosCng  
    will  do  ...  
    except  during  the  
    month  of  the  event
    

    View Slide

33. SaaS
    PaaS
    IaaS
    

    View Slide

34. DB  &  
    webserver?
    Or  specialized  tools?
    

    View Slide

35. m
    ulC-­‐server  setup
    About  a
    

    View Slide

36. m
    ulC-­‐server  setup
    About  a
    Loadbalancing
    

    View Slide

37. m
    ulC-­‐server  setup
    About  a
    What  about  uploaded  files?
    

    View Slide

38. m
    ulC-­‐server  setup
    About  a
    What  about  sessions?
    

    View Slide

39. m
    ulC-­‐server  setup
    About  a
    Database  replicaCon
    

    View Slide

40. Impact
    

    View Slide

41. rt_sigaction(SIGQUIT,  {0x7f11489295f0,  [QUIT],  SA_RESTORER|SA_INTERRUPT,  0x7f114d583f
    rt_sigprocmask(SIG_UNBLOCK,  [QUIT],  NULL,  8)  =  0
    rt_sigprocmask(SIG_BLOCK,  [TERM],  NULL,  8)  =  0
    rt_sigaction(SIGTERM,  {0x7f11489295f0,  [TERM],  SA_RESTORER|SA_INTERRUPT,  0x7f114d583f
    rt_sigprocmask(SIG_UNBLOCK,  [TERM],  NULL,  8)  =  0
    rt_sigprocmask(SIG_BLOCK,  [XCPU],  NULL,  8)  =  0
    rt_sigaction(SIGXCPU,  {0x7f11489295f0,  [XCPU],  SA_RESTORER|SA_INTERRUPT,  0x7f114d583f
    rt_sigprocmask(SIG_UNBLOCK,  [XCPU],  NULL,  8)  =  0
    rt_sigprocmask(SIG_BLOCK,  [XFSZ],  NULL,  8)  =  0
    rt_sigaction(SIGXFSZ,  {0x7f11489295f0,  [XFSZ],  SA_RESTORER|SA_INTERRUPT,  0x7f114d583f
    rt_sigprocmask(SIG_UNBLOCK,  [XFSZ],  NULL,  8)  =  0
    open("/proc/meminfo",  O_RDONLY)                  =  3
    fstat(3,  {st_mode=S_IFREG|0444,  st_size=0,  ...})  =  0
    mmap(NULL,  4096,  PROT_READ|PROT_WRITE,  MAP_PRIVATE|MAP_ANONYMOUS,  -­‐1,  0)  =  0x7f115022
    read(3,  "MemTotal:                2097152  kB\nMemFr"...,  1024)  =  560
    close(3)                                                                =  0
    munmap(0x7f1150225000,  4096)                        =  0
    getrlimit(RLIMIT_NOFILE,  {rlim_cur=1024,  rlim_max=1024})  =  0
    brk(0x1ae1000)                                                    =  0x1ae1000
    brk(0x1b02000)                                                    =  0x1b02000
    socket(PF_FILE,  SOCK_STREAM,  0)                  =  3
    fcntl(3,  F_SETFL,  O_RDWR|O_NONBLOCK)        =  0
    connect(3,  {sa_family=AF_FILE,  path="/var/run/nscd/socket"...},  110)  =  -­‐1  ECONNREFUSE
    close(3)                                                                =  0
    socket(PF_FILE,  SOCK_STREAM,  0)                  =  3
    fcntl(3,  F_SETFL,  O_RDWR|O_NONBLOCK)        =  0
    connect(3,  {sa_family=AF_FILE,  path="/var/run/nscd/socket"...},  110)  =  -­‐1  ECONNREFUSE
    close(3)                                                                =  0
    open("/etc/services",  O_RDONLY|O_CLOEXEC)  =  3
    fstat(3,  {st_mode=S_IFREG|0644,  st_size=18480,  ...})  =  0
    System
    calls
    

    View Slide

42. rt_sigaction(SIGQUIT,  {0x7f11489295f0,  [QUIT],  SA_RESTORER|SA_INTERRUPT,  0x7f114d583f
    rt_sigprocmask(SIG_UNBLOCK,  [QUIT],  NULL,  8)  =  0
    rt_sigprocmask(SIG_BLOCK,  [TERM],  NULL,  8)  =  0
    rt_sigaction(SIGTERM,  {0x7f11489295f0,  [TERM],  SA_RESTORER|SA_INTERRUPT,  0x7f114d583f
    rt_sigprocmask(SIG_UNBLOCK,  [TERM],  NULL,  8)  =  0
    rt_sigprocmask(SIG_BLOCK,  [XCPU],  NULL,  8)  =  0
    rt_sigaction(SIGXCPU,  {0x7f11489295f0,  [XCPU],  SA_RESTORER|SA_INTERRUPT,  0x7f114d583f
    rt_sigprocmask(SIG_UNBLOCK,  [XCPU],  NULL,  8)  =  0
    rt_sigprocmask(SIG_BLOCK,  [XFSZ],  NULL,  8)  =  0
    rt_sigaction(SIGXFSZ,  {0x7f11489295f0,  [XFSZ],  SA_RESTORER|SA_INTERRUPT,  0x7f114d583f
    rt_sigprocmask(SIG_UNBLOCK,  [XFSZ],  NULL,  8)  =  0
    open("/proc/meminfo",  O_RDONLY)                  =  3
    fstat(3,  {st_mode=S_IFREG|0444,  st_size=0,  ...})  =  0
    mmap(NULL,  4096,  PROT_READ|PROT_WRITE,  MAP_PRIVATE|MAP_ANONYMOUS,  -­‐1,  0)  =  0x7f115022
    read(3,  "MemTotal:                2097152  kB\nMemFr"...,  1024)  =  560
    close(3)                                                                =  0
    munmap(0x7f1150225000,  4096)                        =  0
    getrlimit(RLIMIT_NOFILE,  {rlim_cur=1024,  rlim_max=1024})  =  0
    brk(0x1ae1000)                                                    =  0x1ae1000
    brk(0x1b02000)                                                    =  0x1b02000
    socket(PF_FILE,  SOCK_STREAM,  0)                  =  3
    fcntl(3,  F_SETFL,  O_RDWR|O_NONBLOCK)        =  0
    connect(3,  {sa_family=AF_FILE,  path="/var/run/nscd/socket"...},  110)  =  -­‐1  ECONNREFUSE
    close(3)                                                                =  0
    socket(PF_FILE,  SOCK_STREAM,  0)                  =  3
    fcntl(3,  F_SETFL,  O_RDWR|O_NONBLOCK)        =  0
    connect(3,  {sa_family=AF_FILE,  path="/var/run/nscd/socket"...},  110)  =  -­‐1  ECONNREFUSE
    close(3)                                                                =  0
    open("/etc/services",  O_RDONLY|O_CLOEXEC)  =  3
    fstat(3,  {st_mode=S_IFREG|0644,  st_size=18480,  ...})  =  0
    

    View Slide

43. CPU
    RAM
    IO
    

    View Slide

44. Modular  
    architecture
    

    View Slide

45. Connection  
    handling
    strategy
    

    View Slide

46. Path  
    lookups
    

    View Slide

47. Caching
    

    View Slide

48. Is  caching  
    the  
    answer?
    

    View Slide

49. Frontend Backend
    Data  
    cache
    Opcode  
    cache
    Page  
    cache
    APC
    Memcached
    SQLite
    File
    Redis
    +more
    +more
    

    View Slide

50. What?
    When  not?
    How  long?
    CompaCbility
    Limited  size
    

    View Slide

51. APC
    

    View Slide

52. Do  you  have  a  
    data  access  
    strategy?
    

    View Slide

53. Distributed  
    cache
    

    View Slide

54. View Slide

55. memcached.sess_prefix = "memc.sess.key."
    session.save_path="127.0.0.1:11211"
    session.save_handler = memcached
    Also  
    for  session  
    clustering
    

    View Slide

56. View Slide

57. Reverse  
    (caching)  
    proxies
    

    View Slide

58. Protects  your  
    servers
    

    View Slide

59. Put  ‘em  before  
    your  
    webserver
    

    View Slide

60. They  honor  
    your  caching  
    headers
    

    View Slide

61. View Slide

62. View Slide

63. Content
    Delivery
    Network
    

    View Slide

64. View Slide

65. View Slide

66. I like Varnish
    

    View Slide

67. Exceptional  
    performance
    

    View Slide

68. Open  source  
    software
    

    View Slide

69. Varnish  
    Configuration  
    Language
    

    View Slide

70. Supports  
    ESI
    

    View Slide

71. Can  handle  
    purging  &  
    banning
    

    View Slide

72. backend default {
    .host = "127.0.0.1";
    .port = "8080";
    }
    acl purge {
    "localhost";
    "127.0.0.1";
    "some.host.name.com";
    "1.2.3.4";
    }
    

    View Slide

73. sub vcl_recv {
    if (req.request == "PURGE") {
    if (!client.ip ~ purge) {
    error 405 "Not allowed: " + client.ip;
    }
    return(lookup);
    }
    if (req.request != "GET" &&
    req.request != "HEAD" &&
    req.request != "PUT" &&
    req.request != "POST" &&
    req.request != "TRACE" &&
    req.request != "OPTIONS" &&
    req.request != "DELETE") {
    return (pipe);
    }
    

    View Slide

74. if (req.request != "GET" && req.request != "HEAD") {
    return (pass);
    }
    if (req.http.Accept-Encoding) {
    if (req.url ~ "\.(gif|jpg|jpeg|swf|flv|mp3|mp4|
    pdf|ico|png|gz|tgz|bz2)(\?.*|)$") {
    remove req.http.Accept-Encoding;
    } else if (req.http.Accept-Encoding ~ "gzip") {
    set req.http.Accept-Encoding = "gzip";
    } else if (req.http.Accept-Encoding ~ "deflate"){
    set req.http.Accept-Encoding = "deflate";
    } else {
    remove req.http.Accept-Encoding;
    }
    }
    

    View Slide

75. if (req.url ~ "\.(gif|jpg|jpeg|swf|css|js|flv|mp3|
    mp4|pdf|ico|png)(\?.*|)$") {
    unset req.http.cookie;
    set req.url = regsub(req.url, "\?.*$", "");
    return (lookup);
    }
    if (req.url ~ "\?(utm_(campaign|medium|source|term)|
    adParams|client|cx|eid|fbid|feed|ref(id|src)?|v(er|
    iew))=") {
    set req.url = regsub(req.url, "\?.*$", "");
    }
    if (req.http.cookie ~ "^ *$") {
    unset req.http.cookie;
    }
    

    View Slide

76. sub vcl_fetch {
    if (req.url ~ "wp-(login|admin)" || req.url ~
    "preview=true" || req.url ~ "xmlrpc.php") {
    return (hit_for_pass);
    }
    if ( (!(req.url ~ "(wp-(login|admin)|login)"))) {
    unset beresp.http.set-cookie;
    set beresp.ttl = 1h;
    }
    if (req.url ~ "\.(gif|jpg|jpeg|swf|css|js|flv|mp3|
    mp4|pdf|ico|png)(\?.*|)$") {
    set beresp.ttl = 365d;
    }
    }
    

    View Slide

77. sub vcl_deliver {
    if (obj.hits > 0) {
    set resp.http.X-Cache = "HIT";
    } else {
    set resp.http.X-Cache = "MISS";
    }
    }
    sub vcl_hit {
    if (req.request == "PURGE") {
    purge;
    set obj.ttl = 0s;
    error 200 "OK";
    }
    }
    sub vcl_miss {
    if (req.request == "PURGE") {
    purge;
    error 200 "OK";
    }
    }
    

    View Slide

78. Webserver    
    performance
    

    View Slide

79. Apache  vs  Nginx
    

    View Slide

80. Prefork  or  worker  mode
    Async,  event-­‐driven  workers
    Apache
    Nginx
    

    View Slide

81. Evented  non-­‐blocking  I/O
    We’ll  call  
    you  back!
    

    View Slide

82. Blocking
    var  results  =  db.query("SELECT  *  FROM  tablename");
    //use  results  
    Non-­‐blocking
    db.query("SELECT  *  FROM  tablename"  ,  function  (results)  {  
    //use  results  
    });  
    //do  something  else  
    

    View Slide

83. Example
    var  http  =  require('http');
    http.createServer(function  (req,  res)  {
       res.writeHead(200,  {'Content-­‐Type':  'text/plain'});
       res.end('Hello  World\n');
    }).listen(1337,  '127.0.0.1');
    console.log('Server  running  at  http://127.0.0.1:1337/');
    

    View Slide

84. We  can  do  it  in  PHP  too
    

    View Slide

85. Example
    require 'vendor/autoload.php';
    $app = function ($request, $response) {
    $response->writeHead(200, array('Content-
    Type' => 'text/plain'));
    $response->end("Hello World\n");
    };
    $loop = React\EventLoop\Factory::create();
    $socket = new React\Socket\Server($loop);
    $http = new React\Http\Server($socket, $loop);
    $http->on('request', $app);
    echo "Server running at http://127.0.0.1:1337\n";
    $socket->listen(1337);
    $loop->run();
    

    View Slide

86. Real  /me  data
    Ajax
    Websockets
    

    View Slide

87. Websockets
    var conn = new
    WebSocket('ws://
    some.host');
    conn.onmessage =
    function(e) {
    console.log(e.data);
    };
    conn.send('Hello World!');
    Client
    

    View Slide

88. Websockets
    Server
    Ratchet:  
    websocket  
    framework  
    on  top  of  
    React
    

    View Slide

89. namespace MyApp;
    use Ratchet\MessageComponentInterface;
    use Ratchet\ConnectionInterface;
    class Chat implements MessageComponentInterface {
    protected $clients;
    public function __construct() {
    $this->clients = new \SplObjectStorage;
    }
    public function onOpen(ConnectionInterface $conn) {
    $this->clients->attach($conn);
    }
    public function onMessage(ConnectionInterface $from, $msg) {
    foreach ($this->clients as $client) {
    if ($from !== $client) {
    $client->send('Response: '.$msg);
    }
    }
    }
    public function onClose(ConnectionInterface $conn) {
    $this->clients->detach($conn);
    }
    public function onError(ConnectionInterface $conn, \Exception $e) {
    echo "An error has occurred: {$e->getMessage()}\n";
    $conn->close();
    }
    }
    

    View Slide

90. require __DIR__.'/vendor/autoload.php';
    require 'chat.php';
    use Ratchet\WebSocket\WsServer;
    use Ratchet\Server\IoServer;
    use MyApp\Chat;
    $ws = new WsServer(new Chat());
    $ws->disableVersion(0);
    $server = IoServer::factory($ws,80,'1.2.3.4');
    $server->run();
    

    View Slide

91. FPM
    hJp://www.php.net/manual/en/install.fpm.configuraOon.php
    

    View Slide

92. [www]
    listen = 9000
    user = www-data
    group = www-data
    pm = dynamic
    pm.max_children = 10
    pm.start_servers = 4
    pm.max_spare_servers = 6
    pm.max_requests = 500
    php_admin_value[sendmail_path] = /usr/sbin/sendmail -t -i -f
    www@my.domain.com
    php_flag[display_errors] = off
    php_admin_value[error_log] = /var/log/fpm-php.www.log
    php_admin_flag[log_errors] = on
    php_admin_value[memory_limit] = 32M
    

    View Slide

93. upstream fpm {
    server php1.server.com:9000;
    server php2.server.com:9000;
    }
    upstream memcached {
    server memcached1.server.com:11211;
    server memcached2.server.com:11211;
    }
    server {
    root /var/www;
    index index.php index.html index.htm;
    server_name nginx.server.com;
    location / {
    try_files $uri $uri/ /index.php;
    }
    location @php {
    fastcgi_pass fpm;
    fastcgi_index index.php;
    include fastcgi_params;
    }
    

    View Slide

94. location ~ \.php$ {
    set $memcached_key $request_uri;
    memcached_pass memcached;
    memcached_next_upstream not_found;
    default_type text/html;
    error_page 404 405 502 = @php;
    }
    location ~* \.(?:ico|css|js|gif|jpe?g|png)$ {
    expires max;
    add_header Pragma public;
    add_header Cache-Control "public, must-
    revalidate, proxy-revalidate";
    }
    }
    

    View Slide

95. Offloading
    Resources
    

    View Slide

96. fastcgi_finish_request();
    

    View Slide

97. Disconnects  from  webserver
    ConCnues  async  processing
    

    View Slide

98. View Slide

99. View Slide

100. # Reverse Worker Code
     $worker= new GearmanWorker();
     $worker->addServer();
     $worker->addFunction("reverse",
     "my_reverse_function");
     while ($worker->work());
     function my_reverse_function($job)
     {
     return strrev($job->workload());
     }
     # Reverse Client Code
     $client= new GearmanClient();
     $client->addServer();
     print $client->do("reverse", "Hello World!");
     

     View Slide

101. # Reverse Client Code
     $client= new GearmanClient();
     $client->addServer();
     print $client->do("reverse", "Hello World!");
     

     View Slide

102. Don’t  be  clueless
     

     View Slide

103. Apply  the  tricks
     

     View Slide

104. And  don’t  forget  to  refactor  aPerwards
     

     View Slide

105. And  then  relax  ...  and  be  happy
     

     View Slide

106. View Slide