8 <object id="TOC" data="graphical-toc.svg" width="100%" type="image/svg+xml"></object>
10 window.onload=function() { // Wait for the SVG to be loaded before changing it
11 var elem=document.querySelector("#TOC").contentDocument.getElementById("ActorBox")
12 elem.style="opacity:0.93999999;fill:#ff0000;fill-opacity:0.1;stroke:#000000;stroke-width:0.35277778;stroke-linecap:round;stroke-linejoin:round;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1";
18 The S4U interface (SimGrid for you) mixes the full power of SimGrid
19 with the full power of C++. This is the preferred interface to describe
20 abstract algorithms in the domains of Cloud, P2P, HPC, IoT, and similar
23 Since v3.20 (June 2018), S4U is definitely the way to go for long-term
24 projects. It is feature complete, but may still evolve slightly in the
25 future releases. It can already be used to do everything that can be
26 done in SimGrid, but you may have to adapt your code in future
27 releases. When this happens, compiling your code will produce
28 deprecation warnings for 4 releases (one year) before the removal of
30 If you want an API that will never ever evolve in the future, you
31 should use the deprecated MSG API instead.
36 A typical SimGrid simulation is composed of several |API_s4u_Actors|_, that
37 execute user-provided functions. The actors have to explicitly use the
38 S4U interface to express their :ref:`computation <API_s4u_Exec>`,
39 :ref:`communication <API_s4u_Comm>`, :ref:`disk usage <API_s4u_Io>`,
40 and other |API_s4u_Activities|_, so that they get reflected within the
41 simulator. These activities take place on resources such as |API_s4u_Hosts|_,
42 |API_s4u_Links|_ and |API_s4u_Disks|_. SimGrid predicts the time taken by each
43 activity and orchestrates the actors accordingly, waiting for the
44 completion of these activities.
47 When **communicating**, data is not directly sent to other actors but
48 posted onto a |API_s4u_Mailbox|_ that serves as a rendez-vous point between
49 communicating actors. This means that you don't need to know who you
50 are talking to, you just put your communication `Put` request in a
51 mailbox, and it will be matched with a complementary `Get`
52 request. Alternatively, actors can interact through **classical
53 synchronization mechanisms** such as |API_s4u_Barrier|_, |API_s4u_Semaphore|_,
54 |API_s4u_Mutex|_ and |API_s4u_ConditionVariable|_.
56 Each actor is located on a simulated |API_s4u_Host|_. Each host is located
57 itself in a |API_s4u_NetZone|_, that knows the networking path between one
58 resource to another. Each NetZone is included in another one, forming
59 a tree of NetZones which root zone contains the whole platform. The
60 actors can also be located on a |API_s4U_VirtualMachine|_ that may
61 restrict the activities it contains to a limited amount of cores.
62 Virtual machines can also be migrated between hosts.
64 The :ref:`simgrid::s4u::this_actor <API_s4u_this_actor>` namespace
65 provides many helper functions to simplify the code of actors.
69 - :ref:`class s4u::Actor <API_s4u_Actor>`:
70 Active entities executing your application.
71 - :ref:`class s4u::Engine <API_s4u_Engine>`
72 Simulation engine (singleton).
73 - :ref:`class s4u::Mailbox <API_s4u_Mailbox>`
74 Communication rendez-vous.
76 - **Platform Elements**
78 - :ref:`class s4u::Disk <API_s4u_Disk>`
79 Resource on which actors can write and read data.
80 - :ref:`class s4u::Host <API_s4u_Host>`:
81 Actor location, providing computational power.
82 - :ref:`class s4u::Link <API_s4u_Link>`
83 Interconnecting hosts.
84 - :ref:`class s4u::NetZone <API_s4u_NetZone>`:
85 Sub-region of the platform, containing resources (Hosts, Links, etc).
86 - :ref:`class s4u::VirtualMachine <API_s4u_VirtualMachine>`:
87 Execution containers that can be moved between Hosts.
89 - **Activities** (:ref:`class s4u::Activity <API_s4u_Activity>`):
90 The things that actors can do on resources
92 - :ref:`class s4u::Comm <API_s4u_Comm>`
93 Communication activity, started on Mailboxes and consuming links.
94 - :ref:`class s4u::Exec <API_s4u_Exec>`
95 Computation activity, started on Host and consuming CPU resources.
96 - :ref:`class s4u::Io <API_s4u_Io>`
97 I/O activity, started on and consumming disks.
99 - **Synchronization Mechanisms**: Classical IPC that actors can use
101 - :ref:`class s4u::Barrier <API_s4u_Barrier>`
102 - :ref:`class s4u::ConditionVariable <API_s4u_ConditionVariable>`
103 - :ref:`class s4u::Mutex <API_s4u_Mutex>`
104 - :ref:`class s4u::Semaphore <API_s4u_Semaphore>`
107 .. |API_s4u_Actors| replace:: **Actors**
108 .. _API_s4u_Actors: #s4u-actor
110 .. |API_s4u_Activities| replace:: **Activities**
111 .. _API_s4u_Activities: #s4u-activity
113 .. |API_s4u_Hosts| replace:: **Hosts**
114 .. _API_s4u_Hosts: #s4u-host
116 .. |API_s4u_Links| replace:: **Links**
117 .. _API_s4u_Links: #s4u-link
119 .. |API_s4u_Disks| replace:: **Disks**
120 .. _API_s4u_Disks: #s4u-disk
122 .. |API_s4u_VirtualMachine| replace:: **VirtualMachines**
124 .. |API_s4u_Host| replace:: **Host**
126 .. |API_s4u_Mailbox| replace:: **Mailbox**
128 .. |API_s4u_Mailboxes| replace:: **Mailboxes**
129 .. _API_s4u_Mailboxes: #s4u-mailbox
131 .. |API_s4u_NetZone| replace:: **NetZone**
133 .. |API_s4u_Barrier| replace:: **Barrier**
135 .. |API_s4u_Semaphore| replace:: **Semaphore**
137 .. |API_s4u_ConditionVariable| replace:: **ConditionVariable**
139 .. |API_s4u_Mutex| replace:: **Mutex**
143 .. include:: ../../examples/s4u/README.rst
148 Activities represent the actions that consume a resource, such as a
149 :ref:`s4u::Comm <API_s4u_Comm>` that consumes the *transmitting power* of
150 :ref:`s4u::Link <API_s4u_Link>` resources.
152 =======================
153 Asynchronous Activities
154 =======================
156 Every activity can be either **blocking** or **asynchronous**. For
157 example, :cpp:func:`s4u::Mailbox::put() <simgrid::s4u::Mailbox::put>`
158 and :cpp:func:`s4u::Mailbox::get() <simgrid::s4u::Mailbox::get>`
159 create blocking communications: the actor is blocked until the
160 completion of that communication. Asynchronous communications do not
161 block the actor during their execution but progress on their own.
163 Once your asynchronous activity is started, you can test for its
164 completion using :cpp:func:`s4u::Activity::test() <simgrid::s4u::Activity::test>`.
165 This function returns ``true`` if the activity completed already.
166 You can also use :cpp:func:`s4u::Activity::wait() <simgrid::s4u::Activity::wait>`
167 to block until the completion of the activity. To wait for at most a given amount of time,
168 use :cpp:func:`s4u::Activity::wait_for() <simgrid::s4u::Activity::wait_for>`.
169 Finally, to wait at most until a specified time limit, use
170 :cpp:func:`s4u::Activity::wait_until() <simgrid::s4u::Activity::wait_until>`.
174 wait_for and wait_until are currently not implemented for Exec and Io activities.
176 Every kind of activity can be asynchronous:
178 - :ref:`s4u::CommPtr <API_s4u_Comm>` are created with
179 :cpp:func:`s4u::Mailbox::put_async() <simgrid::s4u::Mailbox::put_async>` and
180 :cpp:func:`s4u::Mailbox::get_async() <simgrid::s4u::Mailbox::get_async>`.
181 - :ref:`s4u::IoPtr <API_s4u_Io>` are created with
182 :cpp:func:`s4u::Disk::read_async() <simgrid::s4u::Disk::read_async>` and
183 :cpp:func:`s4u::Disk::write_async() <simgrid::s4u::Disk::write_async>`.
184 - :ref:`s4u::ExecPtr <API_s4u_Exec>` are created with
185 :cpp:func:`s4u::Host::exec_async() <simgrid::s4u::Host::exec_async>`.
186 - In the future, it will become possible to have asynchronous IPC
187 such as asynchronous mutex lock requests.
189 The following example shows how to have several concurrent
190 communications ongoing. First, you have to declare a vector in which
191 we will store the ongoing communications. It is also useful to have a
194 .. literalinclude:: ../../examples/s4u/async-waitall/s4u-async-waitall.cpp
196 :start-after: init-begin
197 :end-before: init-end
200 Then, you start all the communications that should occur concurrently with
201 :cpp:func:`s4u::Mailbox::put_async() <simgrid::s4u::Mailbox::put_async>`.
202 Finally, the actor waits for the completion of all of them at once
204 :cpp:func:`s4u::Comm::wait_all() <simgrid::s4u::Comm::wait_all>`.
206 .. literalinclude:: ../../examples/s4u/async-waitall/s4u-async-waitall.cpp
208 :start-after: put-begin
213 =====================
214 Activities Life cycle
215 =====================
217 Sometimes, you want to change the setting of an activity before it even starts.
219 .. todo:: write this section
226 Please also refer to the :ref:`API reference for s4u::Mailbox
233 |API_s4u_Mailboxes|_ are rendez-vous points for network communications,
234 similar to URLs on which you could post and retrieve data. Actually,
235 the mailboxes are not involved in the communication once it starts,
236 but only to find the contact with which you want to communicate.
238 They are similar to many common things: The phone number, which allows
239 the caller to find the receiver. The twitter hashtag, which help
240 senders and receivers to find each others. In TCP, the pair
241 ``{host name, host port}`` to which you can connect to find your peer.
242 In HTTP, URLs through which the clients can connect to the servers.
243 In ZeroMQ, the queues are used to match senders and receivers.
245 One big difference with most of these systems is that no actor is the
246 exclusive owner of a mailbox, neither in sending nor in receiving.
247 Many actors can send into and/or receive from the same mailbox. TCP
248 socket ports for example are shared on the sender side but exclusive
249 on the receiver side (only one process can receive from a given socket
250 at a given point of time).
252 A big difference with TCP sockets or MPI communications is that
253 communications do not start right away after a
254 :cpp:func:`Mailbox::put() <simgrid::s4u::Mailbox::put()>`, but wait
255 for the corresponding :cpp:func:`Mailbox::get() <simgrid::s4u::Mailbox::get()>`.
256 You can change this by :ref:`declaring a receiving actor <s4u_receiving_actor>`.
258 A big difference with twitter hashtags is that SimGrid does not
259 offer easy support to broadcast a given message to many
260 receivers. So that would be like a twitter tag where each message
261 is consumed by the first receiver.
263 A big difference with the ZeroMQ queues is that you cannot filter
264 on the data you want to get from the mailbox. To model such settings
265 in SimGrid, you'd have one mailbox per potential topic, and subscribe
266 to each topic individually with a
267 :cpp:func:`get_async() <simgrid::s4u::Mailbox::get_async()>` on each mailbox.
268 Then, use :cpp:func:`Comm::wait_any() <simgrid::s4u::Comm::wait_any()>`
269 to get the first message on any of the mailbox you are subscribed onto.
271 The mailboxes are not located on the network, and you can access
272 them without any latency. The network delay are only related to the
273 location of the sender and receiver once the match between them is
274 done on the mailbox. This is just like the phone number that you
275 can use locally, and the geographical distance only comes into play
276 once you start the communication by dialing this number.
278 =====================
279 How to use Mailboxes?
280 =====================
282 You can retrieve any existing mailbox from its name (which is a
283 unique string, just like a twitter tag). This results in a
284 versatile mechanism that can be used to build many different
287 To model classical socket communications, use "hostname:port" as
288 mailbox names, and make sure that only one actor reads into a given
289 mailbox. This does not make it easy to build a perfectly realistic
290 model of the TCP sockets, but in most cases, this system is too
291 cumbersome for your simulations anyway. You probably want something
292 simpler, that turns our to be easy to build with the mailboxes.
294 Many SimGrid examples use a sort of yellow page system where the
295 mailbox names are the name of the service (such as "worker",
296 "master" or "reducer"). That way, you don't have to know where your
297 peer is located to contact it. You don't even need its name. Its
298 function is enough for that. This also gives you some sort of load
299 balancing for free if more than one actor pulls from the mailbox:
300 the first actor that can deal with the request will handle it.
302 =========================================
303 How are put() and get() requests matched?
304 =========================================
306 The matching algorithm simple: first come, first serve. When a new
307 send arrives, it matches the oldest enqueued receive. If no receive is
308 currently enqueued, then the incoming send is enqueued. As you can
309 see, the mailbox cannot contain both send and receive requests: all
310 enqueued requests must be of the same sort.
312 .. _s4u_receiving_actor:
314 ===========================
315 Declaring a Receiving Actor
316 ===========================
318 The last twist is that by default in the simulator, the data starts
319 to be exchanged only when both the sender and the receiver are
320 announced (it waits until both :cpp:func:`put() <simgrid::s4u::Mailbox::put()>`
321 and :cpp:func:`get() <simgrid::s4u::Mailbox::get()>` are posted).
322 In TCP, since you establish connections beforehand, the data starts to
323 flow as soon as the sender posts it, even if the receiver did not post
324 its :cpp:func:`recv() <simgrid::s4u::Mailbox::recv()>` yet.
326 To model this in SimGrid, you can declare a specific receiver to a
327 given mailbox (with the function
328 :cpp:func:`set_receiver() <simgrid::s4u::Mailbox::set_receiver()>`).
329 That way, any :cpp:func:`put() <simgrid::s4u::Mailbox::put()>`
330 posted to that mailbox will start as soon as possible, and the data
331 will already be there on the receiver host when the receiver actor
332 posts its :cpp:func:`get() <simgrid::s4u::Mailbox::get()>`
334 Note that being permanent receivers of a mailbox prevents actors to be
335 garbage-collected. If your simulation creates many short-lived actors
336 that marked as permanent receiver, you should call
337 ``mailbox->set_receiver(nullptr)`` by the end of the actors so that their
338 memory gets properly reclaimed. This call should be at the end of the
339 actor's function, not in a on_exit callback.
344 For sake of simplicity, we use `RAII
345 <https://en.wikipedia.org/wiki/Resource_Acquisition_Is_Initialization>`_
346 for many classes in S4U. This is an idiom where resources are automatically
347 managed through the context. Provided that you never manipulate
348 objects of type Foo directly but always FooPtr references (which are
349 defined as `boost::intrusive_ptr
350 <http://www.boost.org/doc/libs/1_61_0/libs/smart_ptr/intrusive_ptr.html>`_
351 <Foo>), you will never have to explicitly release the resource that
352 you use nor to free the memory of unused objects.
353 Here is a little example:
359 simgrid::s4u::MutexPtr mutex = simgrid::s4u::Mutex::create(); // Too bad we cannot use `new`
361 mutex->lock(); // use the mutex as a simple reference
365 } // The mutex gets automatically freed because the only existing reference gets out of scope
367 Note that Mailboxes, Hosts and Links are not handled thought smart
368 pointers (yet?). This means that it is currently impossible to destroy a
369 mailbox or a link. You can still destroy an host (but probably
370 shouldn't), using :cpp:func:`simgrid::s4u::Host::destroy`.
375 .. _API_s4u_this_actor:
377 =========================
378 namespace s4u::this_actor
379 =========================
381 .. doxygennamespace:: simgrid::s4u::this_actor
382 .. _API_s4u_Activity:
388 .. doxygenclass:: simgrid::s4u::Activity
399 .. doxygentypedef:: ActorPtr
401 .. doxygentypedef:: aid_t
403 .. doxygenclass:: simgrid::s4u::Actor
414 .. doxygentypedef:: BarrierPtr
416 .. doxygenclass:: simgrid::s4u::Barrier
427 .. doxygentypedef:: CommPtr
429 .. doxygenclass:: simgrid::s4u::Comm
434 .. _API_s4u_ConditionVariable:
436 ======================
437 s4u::ConditionVariable
438 ======================
440 .. doxygentypedef:: ConditionVariablePtr
442 .. doxygenclass:: simgrid::s4u::ConditionVariable
453 .. doxygenclass:: simgrid::s4u::Disk
464 .. doxygenclass:: simgrid::s4u::Engine
475 .. doxygentypedef:: ExecPtr
477 .. doxygenclass:: simgrid::s4u::Exec
488 .. doxygentypedef:: ExecSeqPtr
490 .. doxygenclass:: simgrid::s4u::ExecSeq
501 .. doxygentypedef:: ExecParPtr
503 .. doxygenclass:: simgrid::s4u::ExecPar
514 .. doxygenclass:: simgrid::s4u::Host
525 .. doxygentypedef:: IoPtr
527 .. doxygenclass:: simgrid::s4u::Io
538 .. doxygenclass:: simgrid::s4u::Link
549 Please also refer to the :ref:`full doc on s4u::Mailbox <s4u_mailbox>`.
551 .. doxygenclass:: simgrid::s4u::Mailbox
562 .. doxygentypedef:: MutexPtr
564 .. doxygenclass:: simgrid::s4u::Mutex
575 .. doxygenclass:: simgrid::s4u::NetZone
580 .. _API_s4u_Semaphore:
586 .. doxygentypedef:: SemaphorePtr
588 .. doxygenclass:: simgrid::s4u::Semaphore
593 .. _API_s4u_VirtualMachine:
599 .. doxygenclass:: simgrid::s4u::VirtualMachine
611 .. doxygenfunction:: simgrid_init
612 .. doxygenfunction:: simgrid_get_clock
613 .. doxygenfunction:: simgrid_load_deployment
614 .. doxygenfunction:: simgrid_load_platform
615 .. doxygenfunction:: simgrid_register_default
616 .. doxygenfunction:: simgrid_register_function
617 .. doxygenfunction:: simgrid_run
623 See also the :ref:`C++ API <API_s4u_ConditionVariable>`.
625 .. doxygenfunction:: sg_cond_init
626 .. doxygenfunction:: sg_cond_notify_all
627 .. doxygenfunction:: sg_cond_notify_one
628 .. doxygenfunction:: sg_cond_wait
629 .. doxygenfunction:: sg_cond_wait_for
634 The Python API is automatically generated with pybind11. It closely mimicks the C++
635 API, to which you should refer for more information.
641 .. automodule:: simgrid.this_actor
648 .. autoclass:: simgrid.Actor
655 .. autoclass:: simgrid.Comm
662 .. autoclass:: simgrid.Engine
669 .. autoclass:: simgrid.Exec
676 .. autoclass:: simgrid.Host
683 .. autoclass:: simgrid.Mailbox