syn_benchmark.erl 9.0 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230
  1. %% ==========================================================================================================
  2. %% Syn - A global Process Registry and Process Group manager.
  3. %%
  4. %% The MIT License (MIT)
  5. %%
  6. %% Copyright (c) 2019-2021 Roberto Ostinelli <roberto@ostinelli.net> and Neato Robotics, Inc.
  7. %%
  8. %% Permission is hereby granted, free of charge, to any person obtaining a copy
  9. %% of this software and associated documentation files (the "Software"), to deal
  10. %% in the Software without restriction, including without limitation the rights
  11. %% to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  12. %% copies of the Software, and to permit persons to whom the Software is
  13. %% furnished to do so, subject to the following conditions:
  14. %%
  15. %% The above copyright notice and this permission notice shall be included in
  16. %% all copies or substantial portions of the Software.
  17. %%
  18. %% THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  19. %% IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  20. %% FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  21. %% AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  22. %% LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  23. %% OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  24. %% THE SOFTWARE.
  25. %% ==========================================================================================================
  26. -module(syn_benchmark).
  27. %% API
  28. -compile([export_all]).
  29. %% ===================================================================
  30. %% API
  31. %% ===================================================================
  32. %% example run: `PROCESS_COUNT=100000 NODES_COUNT=2 make bench`
  33. start() ->
  34. %% init
  35. SlavesCount = list_to_integer(os:getenv("NODES_COUNT", "1")),
  36. ProcessCount = list_to_integer(os:getenv("PROCESS_COUNT", "100000")),
  37. ProcessesPerNode = round(ProcessCount / SlavesCount),
  38. io:format("-----> Starting benchmark on ~w nodes (~w slaves) (for ~w processes total (~w / slave node)~n",
  39. [SlavesCount + 1, SlavesCount, ProcessCount, ProcessesPerNode]
  40. ),
  41. %% start nodes
  42. NodesInfo = lists:foldl(fun(I, Acc) ->
  43. %% start slave
  44. CountBin = integer_to_binary(I),
  45. NodeShortName = binary_to_atom(<<"slave_", CountBin/binary>>),
  46. {ok, Node} = ct_slave:start(NodeShortName, [
  47. {boot_timeout, 10},
  48. {monitor_master, true}
  49. ]),
  50. %% add code path
  51. CodePath = code:get_path(),
  52. true = rpc:call(Node, code, set_path, [CodePath]),
  53. %% start syn
  54. rpc:call(Node, syn, start, []),
  55. %% gather data
  56. FromName = (I - 1) * ProcessesPerNode + 1,
  57. ToName = FromName + ProcessesPerNode - 1,
  58. %% fold
  59. [{Node, FromName, ToName} | Acc]
  60. end, [], lists:seq(1, SlavesCount)),
  61. %% start syn locally
  62. ok = syn:start(),
  63. timer:sleep(1000),
  64. CollectorPid = self(),
  65. %% start processes
  66. PidsMap = lists:foldl(fun({Node, _FromName, _ToName}, Acc) ->
  67. Pids = rpc:call(Node, ?MODULE, start_processes, [ProcessesPerNode]),
  68. maps:put(Node, Pids, Acc)
  69. end, #{}, NodesInfo),
  70. %% {ok, P} = eprof:start(),
  71. %% eprof:start_profiling(erlang:processes() -- [P]),
  72. %% start registration
  73. lists:foreach(fun({Node, FromName, _ToName}) ->
  74. Pids = maps:get(Node, PidsMap),
  75. rpc:cast(Node, ?MODULE, register_on_node, [CollectorPid, FromName, Pids])
  76. end, NodesInfo),
  77. %% wait
  78. RegRemoteNodesTimes = wait_from_all_remote_nodes(nodes(), []),
  79. io:format("----> Remote registration times:~n"),
  80. io:format(" --> MIN: ~p secs.~n", [lists:min(RegRemoteNodesTimes)]),
  81. io:format(" --> MAX: ~p secs.~n", [lists:max(RegRemoteNodesTimes)]),
  82. {RegPropagationTimeMs, _} = timer:tc(?MODULE, wait_registration_propagation, [NodesInfo]),
  83. RegPropagationTime = RegPropagationTimeMs / 1000000,
  84. io:format("----> Eventual additional time to propagate all to master: ~p secs.~n", [RegPropagationTime]),
  85. %% sum
  86. RegTakenTime = (lists:max(RegRemoteNodesTimes) + RegPropagationTime),
  87. RegistrationRate = ProcessCount / RegTakenTime,
  88. io:format("====> Registeration rate (with propagation): ~p/sec.~n~n", [RegistrationRate]),
  89. timer:sleep(1000),
  90. %% start unregistration
  91. lists:foreach(fun({Node, FromName, ToName}) ->
  92. rpc:cast(Node, ?MODULE, unregister_on_node, [CollectorPid, FromName, ToName])
  93. end, NodesInfo),
  94. %% wait
  95. UnregRemoteNodesTimes = wait_from_all_remote_nodes(nodes(), []),
  96. io:format("----> Remote unregistration times:~n"),
  97. io:format(" --> MIN: ~p secs.~n", [lists:min(UnregRemoteNodesTimes)]),
  98. io:format(" --> MAX: ~p secs.~n", [lists:max(UnregRemoteNodesTimes)]),
  99. {UnregPropagationTimeMs, _} = timer:tc(?MODULE, wait_unregistration_propagation, [NodesInfo]),
  100. UnregPropagationTime = UnregPropagationTimeMs / 1000000,
  101. io:format("----> Eventual additional time to propagate all to master: ~p secs.~n", [UnregPropagationTime]),
  102. %% sum
  103. UnregTakenTime = (lists:max(UnregRemoteNodesTimes) + UnregPropagationTime),
  104. UnregistrationRate = ProcessCount / UnregTakenTime,
  105. io:format("====> Unregisteration rate (with propagation): ~p/sec.~n~n", [UnregistrationRate]),
  106. %% start re-registration
  107. lists:foreach(fun({Node, FromName, _ToName}) ->
  108. Pids = maps:get(Node, PidsMap),
  109. rpc:cast(Node, ?MODULE, register_on_node, [CollectorPid, FromName, Pids])
  110. end, NodesInfo),
  111. %% wait
  112. ReRegRemoteNodesTimes = wait_from_all_remote_nodes(nodes(), []),
  113. io:format("----> Remote re-registration times:~n"),
  114. io:format(" --> MIN: ~p secs.~n", [lists:min(ReRegRemoteNodesTimes)]),
  115. io:format(" --> MAX: ~p secs.~n", [lists:max(ReRegRemoteNodesTimes)]),
  116. {ReRegPropagationTimeMs, _} = timer:tc(?MODULE, wait_registration_propagation, [NodesInfo]),
  117. ReRegPropagationTime = ReRegPropagationTimeMs / 1000000,
  118. io:format("----> Eventual additional time to propagate all to master: ~p secs.~n", [ReRegPropagationTime]),
  119. %% sum
  120. ReRegTakenTime = (lists:max(ReRegRemoteNodesTimes) + ReRegPropagationTime),
  121. ReRegistrationRate = ProcessCount / ReRegTakenTime,
  122. io:format("====> Re-registeration rate (with propagation): ~p/sec.~n~n", [ReRegistrationRate]),
  123. %% kill all processes
  124. maps:foreach(fun(_Node, Pids) ->
  125. lists:foreach(fun(Pid) -> exit(Pid, kill) end, Pids)
  126. end, PidsMap),
  127. %% wait all unregistered
  128. {KillPropagationTimeMs, _} = timer:tc(?MODULE, wait_unregistration_propagation, [NodesInfo]),
  129. KillPropagationTime = KillPropagationTimeMs / 1000000,
  130. io:format("----> Time to propagate killed process to to master: ~p secs.~n", [KillPropagationTime]),
  131. KillRate = ProcessCount / KillPropagationTime,
  132. io:format("====> Unregistered after kill rate (with propagation): ~p/sec.~n~n", [KillRate]),
  133. %% eprof:stop_profiling(),
  134. %% eprof:analyze(total),
  135. %% stop node
  136. init:stop().
  137. register_on_node(CollectorPid, FromName, Pids) ->
  138. {TimeMs, _} = timer:tc(?MODULE, do_register_on_node, [FromName, Pids]),
  139. Time = TimeMs / 1000000,
  140. io:format("----> Registered on node ~p on ~p secs.~n", [node(), Time]),
  141. CollectorPid ! {done, node(), Time}.
  142. do_register_on_node(_Name, []) -> ok;
  143. do_register_on_node(Name, [Pid | PidsTail]) ->
  144. ok = syn:register(Name, Pid),
  145. do_register_on_node(Name + 1, PidsTail).
  146. unregister_on_node(CollectorPid, FromName, ToName) ->
  147. {TimeMs, _} = timer:tc(?MODULE, do_unregister_on_node, [FromName, ToName]),
  148. Time = TimeMs / 1000000,
  149. io:format("----> Unregistered on node ~p on ~p secs.~n", [node(), Time]),
  150. CollectorPid ! {done, node(), Time}.
  151. do_unregister_on_node(FromName, ToName) when FromName > ToName -> ok;
  152. do_unregister_on_node(Name, ToName) ->
  153. ok = syn:unregister(Name),
  154. do_unregister_on_node(Name + 1, ToName).
  155. start_processes(Count) ->
  156. start_processes(Count, []).
  157. start_processes(0, Pids) ->
  158. Pids;
  159. start_processes(Count, Pids) ->
  160. Pid = spawn(fun process_loop/0),
  161. start_processes(Count - 1, [Pid | Pids]).
  162. process_loop() ->
  163. receive
  164. _ -> ok
  165. end.
  166. wait_from_all_remote_nodes([], Times) -> Times;
  167. wait_from_all_remote_nodes([RemoteNode | Tail], Times) ->
  168. receive
  169. {done, RemoteNode, Time} ->
  170. wait_from_all_remote_nodes(Tail, [Time | Times])
  171. end.
  172. wait_registration_propagation([]) -> ok;
  173. wait_registration_propagation([{_Node, _FromName, ToName} | NodeInfosTail] = NodesInfo) ->
  174. case syn:lookup(ToName) of
  175. undefined ->
  176. timer:sleep(50),
  177. wait_registration_propagation(NodesInfo);
  178. {_Pid, undefined} ->
  179. wait_registration_propagation(NodeInfosTail)
  180. end.
  181. wait_unregistration_propagation([]) -> ok;
  182. wait_unregistration_propagation([{_Node, _FromName, ToName} | NodeInfosTail] = NodesInfo) ->
  183. case syn:lookup(ToName) of
  184. undefined ->
  185. wait_unregistration_propagation(NodeInfosTail);
  186. {_Pid, undefined} ->
  187. timer:sleep(50),
  188. wait_unregistration_propagation(NodesInfo)
  189. end.