1 /*! \page platform Platform Description
5 In order to run any simulation, SimGrid must be provided with three things:
6 something to run (i.e., your code), a description of the platform on which you
7 want to simulate your application and lastly information about the deployment
8 process: Which process should be deployed to which processor/core?
10 For the last two items, there are essentially two possible ways you can provide
11 this information as an input:
12 \li You can program it, either by using the Lua console (
13 \ref MSG_Lua_funct) or, if you're using MSG, some of MSG's platform and
14 deployment functions (\ref msg_simulation). If you want to use this,
15 check the particular documentation. (You can also check the section
16 \ref pf_flexml_bypassing, however, this documentation is deprecated;
17 there is a new, but undocumented, way to do it properly).
18 \li You can use two XML files: one contains the platform description while
19 the other contains the deployment instructions.
21 For more information on SimGrid's deployment features, please refer to
22 the \ref deployment documentation.
24 The platform description may be intricate. This documentation is all
25 about how to write this file: The basic concepts are introduced. Furthermore,
26 advanced options are explained. Additionally, some hints and tips on how to
27 write a good platform description are given.
29 \section pf_overview Some words about XML and DTD
31 We chose to use XML not only because it's extensible but also because many
32 tools (and plugins for existing tools) are available that facilitate editing and
33 validating XML files. Furthermore, libraries that parse XML are often already
34 available and very well tested.
36 The XML checking is done based on the Document Type Definition (DTD) file,
38 <a href="http://simgrid.gforge.inria.fr/simgrid.dtd">http://simgrid.gforge.inria.fr/simgrid.dtd</a>.
40 If you read the DTD, you should notice the following:
41 \li The platform tags contain a version attribute; the current version is 3.
42 This property might be used in the future to provide backwards
44 \li The DTD contains definitions for the two files used by SimGrid (i.e.,
45 platform description and deployment).
47 \section pf_basics Basic concepts
49 Nowadays, the Internet is composed of a bunch of independently managed
50 networks. Within each of those networks, there are entry and exit
51 points (most of the time, you can both enter and exit through the same
52 point); this allows to leave the current network and reach other
53 networks, possibly even in other locations.
54 At the upper level, such a network is called
55 <b>Autonomous System (AS)</b>, while at the lower level it is named
56 sub-network, or LAN (local area network).
57 They are indeed autonomous: routing is defined
58 (within the limits of his network) by the administrator, and so, those
59 networks can operate without a connection to other
60 networks. So-called gateways allow you to go from one network to
61 another, if such a (physical) connection exists. Every node in one network
62 that can be directly reached (i.e., without traversing other nodes) from
63 another network is called a gateway.
64 Each autonomous system consists of equipment such as cables (network links),
65 routers and switches as well as computers.
67 The structure of the SimGrid platform description relies exactly on the same
68 concept as a real-life platform (see above). Every resource (computers,
69 network equipment etc.) belongs to an AS, which can be defined by using the
70 \<AS\> tag. Within an AS, the routing between its elements can be defined
71 abitrarily. There are several modes for routing, and exactly one mode must be
72 selected by specifying the routing attribute in the AS tag:
75 <AS id="AS0" routing="Full">
79 Other supported values for the routing attribute can be found below, Section
82 There is also the ``<route>`` tag; this tag takes two attributes, ``src`` (source)
83 and ``dst`` (destination). Both source and destination must be valid identifiers
84 for routers (these will be introduced later). Contained by the ``<route>`` are
85 network links; these links must be used in order to communicate from the source
86 to the destination specified in the tag. Hence, a route merely describes
87 how to reach a router from another router.
90 More information and (code-)examples can be found in Section \ref pf_rm.
92 An AS can also contain itself one or more AS; this allows you to
93 model the hierarchy of your platform.
95 ### Within each AS, the following types of resources exist:
97 %Resource | Documented in Section | Description
98 --------------- | --------------------- | -----------
99 AS | | Every Autonomous System (AS) may contain one or more AS.
100 host | \ref pf_host | This entity carries out the actual computation. For this reason, it contains processors (with potentially multiple cores).
101 router | \ref pf_router | In SimGrid, routers are used to provide helpful information to routing algorithms. Routers may also act as gateways, connecting several autonomous systems with each other.
102 link | \ref pf_link | In SimGrid, (network)links define a connection between two or potentially even more resources. Every link has a bandwidth and a latency and may potentially experience congestion.
103 cluster | \ref pf_cluster | In SimGrid, clusters were introduced to model large and homogenous environments. They are not really a resource by themselves - technically, they are only a shortcut, as they will internally set up all the hosts, network and routing for you, i.e., using this resource, one can easily setup thousands of hosts and links in a few lines of code. Each cluster is itself an AS.
105 %As it is desirable to interconnect these resources, a routing has to be
106 defined. The AS is supposed to be Autonomous, hence this has to be done at the
107 AS level. The AS handles two different types of entities (<b>host/router</b>
108 and <b>AS</b>). However, the user is responsible to define routes between those resources,
109 otherwise entities will be unconnected and therefore unreachable from other
110 entities. Although several routing algorithms are built into SimGrid (see
111 \ref pf_rm), you might encounter a case where you want to define routes
112 manually (for instance, due to specific requirements of your platform).
114 There are three tags to use:
115 \li <b>ASroute</b>: to define routes between two <b>AS</b>
116 \li <b>route</b>: to define routes between two <b>host/router</b>
117 \li <b>bypassRoute</b>: to define routes between two <b>AS</b> that
118 will bypass default routing (as specified by the ``routing`` attribute
119 supplied to ``<AS>``, see above).
121 Here is an illustration of these concepts:
123 ![An illustration of an AS hierarchy. Here, AS1 contains 5 other AS' who in turn may contain other AS' as well.](AS_hierarchy.png)
124 Circles represent processing units and squares represent network routers. Bold
125 lines represent communication links. AS2 models the core of a national
126 network interconnecting a small flat cluster (AS4) and a larger
127 hierarchical cluster (AS5), a subset of a LAN (AS6), and a set of peers
128 scattered around the world (AS7).
130 \section pf_pftags Resource description
132 \subsection pf_As Platform: The <AS> tag
134 The concept of an AS was already outlined above (Section \ref pf_basics);
135 recall that the AS is so important because it groups other resources (such
136 as routers/hosts) together (in fact, these resources must be contained by
139 Available attributes :
141 Attribute name | Mandatory | Values | Description
142 --------------- | --------- | ------ | -----------
143 id | yes | String | The identifier of an AS; facilitates referring to this AS. ID must be unique.
144 routing | yes | Full\| Floyd\| Dijkstra\| DijkstraCache\| none\| Vivaldi\| Cluster | See Section \ref pf_rm for details.
149 <AS id="AS0" routing="Full">
150 <host id="host1" power="1000000000"/>
151 <host id="host2" power="1000000000"/>
152 <link id="link1" bandwidth="125000000" latency="0.000100"/>
153 <route src="host1" dst="host2"><link_ctn id="link1"/></route>
157 In this example, AS0 contains two hosts (host1 and host2). The route
158 between the hosts goes through link1.
160 \subsection pf_Cr Computing resources: hosts, clusters and peers.
162 \subsubsection pf_host <host/>
164 A <b>host</b> represents a computer/node card. Every host is able to execute
165 code and it can send and receive data to/from other hosts. Most importantly,
166 a host can contain more than 1 core.
170 Attribute name | Mandatory | Values | Description
171 --------------- | --------- | ------ | -----------
172 id | yes | String | The identifier of the host. facilitates referring to this AS.
173 power | yes | double (must be > 0.0) | Computational power of every core of this host in FLOPS. Must be larger than 0.0.
174 core | no | int (Default: 1) | The number of cores of this host. If more than one core is specified, the "power" parameter refers to every core, i.e., the total computational power is no_of_cores*power.<br /> If 6 cores are specified, up to 6 tasks can be executed without sharing the computational power; if more than 6 tasks are executed, computational power will be shared among these tasks. <br /> <b>Warning:</b> Although functional, this model was never scientifically assessed.
175 availability | no | int | <b>Specify if the percentage of power available.</b> (What? TODO)
176 availability_file| no | string | (Relative or absolute) filename to use as input; must contain availability traces for this host. The syntax of this file is defined below. <br /> <b>Note:</b> The filename must be specified with your system's format.
177 state | no | ON\|OFF<br/> (Default: ON) | Is this host running or not?
178 state_file | no | string | Same mechanism as availability_file.<br /> <b>Note:</b> The filename must be specified with your system's format.
179 coordinates | no | string | Must be provided when choosing the Vivaldi, coordinate-based routing model for the AS the host belongs to. More details can be found in the Section \ref pf_P2P_tags.
181 ### Possible children: ###
183 Tag name | Description | Documentation
184 ------------ | ----------- | -------------
185 \<mount/\> | Defines mounting points between some storage resource and the host. | \ref pf_sto_mo
186 \<prop/\> | The prop tag allows you to define additional information on this host following the attribute/value schema. You may want to use it to give information to the tool you use for rendering your simulation, for example. | N/A
191 <host id="host1" power="1000000000"/>
192 <host id="host2" power="1000000000">
193 <prop id="color" value="blue"/>
194 <prop id="rendershape" value="square"/>
199 \anchor pf_host_dynamism
200 ### Expressing dynamism ###
202 SimGrid provides mechanisms to change a hosts' availability over
203 time, using the ``availability_file`` attribute to the ``\<host\>`` tag
204 and a separate text file whose syntax is exemplified below.
206 #### Adding a trace file ####
209 <platform version="1">
210 <host id="bob" power="500000000" availability_file="bob.trace" />
214 #### Example of "bob.trace" file ####
223 Let us begin to explain this example by looking at line 2. (Line 1 will become clear soon).
224 The first column describes points in time, in this case, time 0. The second column
225 describes the relative amount of power this host is able to deliver (relative
226 to the maximum performance specified in the ``\<host\>`` tag). (Clearly, the
227 second column needs to contain values that are not smaller than 0 and not larger than 1).
228 In this example, our host will deliver 500 Mflop/s at time 0, as 500 Mflop/s is the
229 maximum performance of this host. At time 11.0, it will
230 deliver half of its maximum performance, i.e., 250 Mflop/s until time 20.0 when it will
231 will start delivering 80\% of its power. In this example, this amounts to 400 Mflop/s.
233 Since the periodicity in line 1 was set to be 1.0, i.e., 1 timestep, this host will
234 continue to provide 500 Mflop/s from time 21. From time 32 it will provide 250 MFlop/s and so on.
236 ### Changing initial state ###
238 It is also possible to specify whether the host is up or down by setting the
239 ``state`` attribute to either <b>ON</b> (default value) or <b>OFF</b>.
241 #### Example: Expliciting the default value "ON" ####
244 <platform version="1">
245 <host id="bob" power="500000000" state="ON" />
249 If you want this host to be unavailable, simply substitute ON with OFF.
251 ### Expressing churn ###
253 To express the fact that a host can change state over time (as in P2P
254 systems, for instance), it is possible to use a file describing the time
255 at which the host is turned on or off. An example of the content
256 of such a file is presented below.
258 #### Adding a state file ####
261 <platform version="1">
262 <host id="bob" power="500000000" state_file="bob.fail" />
266 #### Example of "bob.fail" file ####
274 A negative value means <b>down</b> (i.e., OFF) while a positive one means <b>up and
275 running</b> (i.e., ON). From time 0.0 to time 1.0, the host is on. At time 1.0, it is
276 turned off and at time 2.0, it is turned on again until time 12 (2.0 plus the
277 periodicity 10.0). It will be turned on again at time 13.0 until time 23.0, and
282 \subsubsection pf_cluster <cluster>
284 ``<cluster />`` represents a machine-cluster. It is most commonly used
285 when one wants to define many hosts and a network quickly. Technically,
286 ``cluster`` is a meta-tag: <b>from the inner SimGrid point of
287 view, a cluster is an AS where some optimized routing is defined</b>.
288 The default inner organization of the cluster is as follow:
294 ____________|__________|_____________ backbone
296 l0| l1| l2| l97| l96 | | l99
302 Here, a set of <b>host</b>s is defined. Each of them has a <b>link</b>
303 to a central backbone (backbone is a link itself, as a link can
304 be used to represent a switch, see the switch / link section
305 below for more details about it). A <b>router</b> allows to connect a
306 <b>cluster</b> to the outside world. Internally,
307 SimGrid treats a cluster as an AS containing all hosts: the router is the default
308 gateway for the cluster.
310 There is an alternative organization, which is as follows:
324 The principle is the same, except that there is no backbone. This representation
325 can be obtained easily: just do not set the bb_* attributes.
328 Attribute name | Mandatory | Values | Description
329 --------------- | --------- | ------ | -----------
330 id | yes | string | The identifier of the cluster. Facilitates referring to this cluster.
331 prefix | yes | string | Each node of the cluster has to have a name. This name will be prefixed with this prefix.
332 suffix | yes | string | Each node of the cluster will be suffixed with this suffix
333 radical | yes | string | Regexp used to generate cluster nodes name. Syntax: "10-20" will give you 11 machines numbered from 10 to 20, "10-20;2" will give you 12 machines, one with the number 2, others numbered as before. The produced number is concatenated between prefix and suffix to form machine names.
334 power | yes | int | Same as the ``power`` attribute of the ``\<host\>`` tag.
335 core | no | int (default: 1) | Same as the ``core`` attribute of the ``\<host\>`` tag.
336 bw | yes | int | Bandwidth for the links between nodes and backbone (if any). <b>See <b>link</b> section for syntax/details.</b>
337 lat | yes | int | Latency for the links between nodes and backbone (if any). See <b>link</b> section for syntax/details.
338 sharing_policy | no | string | Sharing policy for the links between nodes and backbone (if any). See <b>link</b> section for syntax/details.
339 bb_bw | no | int | Bandwidth for backbone (if any). See <b>link</b> section for syntax/details. If bb_bw and bb_lat (see below) attributes are omitted, no backbone is created (alternative cluster architecture <b>described before</b>).
340 bb_lat | no | int | Latency for backbone (if any). See <b>link</b> section for syntax/details. If bb_lat and bb_bw (see above) attributes are omitted, no backbone is created (alternative cluster architecture <b>described before</b>).
341 bb_sharing_policy | no | string | Sharing policy for the backbone (if any). See <b>link</b> section for syntax/details.
342 availability_file | no | string | Allows you to use a file as input for availability. Similar to <b>hosts</b> attribute.
343 state_file | no | string | Allows you to use a file as input for states. Similar to <b>hosts</b> attribute.
344 loopback_bw | no | int | Bandwidth for loopback (if any). See <b>link</b> section for syntax/details. If loopback_bw and loopback_lat (see below) attributes are omitted, no loopback link is created and all intra-node communication will use the main network link of the node. Loopback link is a \ref sharing_policy_fatpipe "\b FATPIPE".
345 loopback_lat | no | int | Latency for loopback (if any). See <b>link</b> section for syntax/details. See loopback_bw for more info.
346 topology | no | FLAT\|TORUS\|FAT_TREE (default: FLAT) | Network topology to use. SimGrid currently supports FLAT (with or without backbone, as described before), <a href="http://en.wikipedia.org/wiki/Torus_interconnect">TORUS </a> and FAT_TREE attributes for this tag.
347 topo_parameters | no | string | Specific parameters to pass for the topology defined in the topology tag. For torus networks, comma-separated list of the number of nodes in each dimension of the torus. For fat trees, refer to \ref AsClusterFatTree "AsClusterFatTree documentation".
351 the router name is defined as the resulting String in the following
355 router_name = prefix + clusterId + _router + suffix;
359 #### Cluster example ####
361 Consider the following two (and independent) uses of the ``cluster`` tag:
364 <cluster id="my_cluster_1" prefix="" suffix="" radical="0-262144"
365 power="1e9" bw="125e6" lat="5E-5"/>
367 <cluster id="my_cluster_2" prefix="c-" suffix=".me" radical="0-99"
368 power="1e9" bw="125e6" lat="5E-5"
369 bb_bw="2.25e9" bb_lat="5E-4"/>
372 The second example creates one router and 100 machines with the following names:
374 c-my_cluster_2_router.me
382 \subsubsection pf_peer The <peer> tag
384 This tag represents a peer, as in Peer-to-Peer (P2P) networks. However, internally,
385 SimGrid transforms a peer into an AS (similar to Cluster). Hence, this tag
386 is virtually only a shortcut that comes with some pre-defined resources
387 and values. These are:
389 \li A tiny AS whose routing type is cluster is created
391 \li Two links: One for download and one for upload. This is
392 convenient to use and simulate stuff under the last mile model (e.g., ADSL peers).
393 \li It connects the two links to the host
394 \li It creates a router (a gateway) that serves as an entry point for this peer zone.
395 This router has coordinates.
399 Attribute name | Mandatory | Values | Description
400 --------------- | --------- | ------ | -----------
401 id | yes | string | The identifier of the peer. Facilitates referring to this peer.
402 power | yes | int | See the description of the ``host`` tag for this attribute
403 bw_in | yes | int | Bandwidth downstream
404 bw_out | yes | int | Bandwidth upstream
405 lat | yes | double | Latency for both up- and downstream, in seconds.
406 coordinates | no | string | Coordinates of the gateway for this peer. Example value: 12.8 14.4 6.4
407 sharing_policy | no | SHARED\|FULLDUPLEX (default: FULLDUPLEX) | Sharing policy for links. See <b>link</b> description for details.
408 availability_file| no | string | Availability file for the peer. Same as host availability file. See <b>host</b> description for details.
409 state_file | no | string | State file for the peer. Same as host state file. See <b>host</b> description for details.
411 Internally, SimGrid transforms any ``\<peer/\>`` construct such as
414 coordinates="12.8 14.4 6.4"
420 into an ``\<AS\>`` (see Sections \ref pf_basics and \ref pf_As). In fact, this example of the ``\<peer/\>`` tag
421 is completely equivalent to the following declaration:
424 <AS id="as_FOO" routing="Cluster">
425 <host id="peer_FOO" power="1.5Gf"/>
426 <link id="link_FOO_UP" bandwidth="2.25GBps" latency="500us"/>
427 <link id="link_FOO_DOWN" bandwidth="2.25GBps" latency="500us"/>
428 <router id="router_FOO" coordinates="25.5 9.4 1.4"/>
429 <host_link id="peer_FOO" up="link_FOO_UP" down="link_FOO_DOWN"/>
434 \subsection pf_ne Network equipments: links and routers
436 There are two tags at all times available to represent network entities and
437 several other tags that are available only in certain contexts.
438 1. ``<link>``: Represents a entity that has a limited bandwidth, a
439 latency, and that can be shared according to TCP way to share this
442 The concept of links in SimGrid may not be intuitive, as links are not limited
443 to connecting (exactly) two entities; in fact, you can have more than two equipments
444 connected to it. (In graph theoretical terms: A link in SimGrid is not an edge,
447 2. ``<router/>``: Represents an entity that a message can be routed
448 to, but that is unable to execute any code. In SimGrid, routers have also
449 no impact on the performance: Routers do not limit any bandwidth nor
450 do they increase latency. As a matter of fact, routers are (almost) ignored
451 by the simulator when the simulation has begun.
453 3. ``<backbone/>``: This tag is only available when the containing AS is
454 used as a cluster (i.e., mode="Cluster")
457 If you want to represent an entity like a switch, you must use ``<link>`` (see section). Routers are used
458 to run some routing algorithm and determine routes (see Section \ref pf_routing for details).
460 \subsubsection pf_router <router/>
462 %As said before, <b>router</b> is used only to give some information
463 for routing algorithms. So, it does not have any attributes except :
467 Attribute name | Mandatory | Values | Description
468 --------------- | --------- | ------ | -----------
469 id | yes | string | The identifier of the router to be used when referring to it.
470 coordinates | yes | string | Must be provided when choosing the Vivaldi, coordinate-based routing model for the AS the router belongs to. More details can be found in the Section \ref pf_P2P_tags.
475 <router id="gw_dc1_horizdist"/>
478 \subsubsection pf_link <link/>
480 Network links can represent one-hop network connections. They are
481 characterized by their id and their bandwidth; links can (but may not) be subject
486 Attribute name | Mandatory | Values | Description
487 --------------- | --------- | ------ | -----------
488 id | yes | string | The identifier of the link to be used when referring to it.
489 bandwidth | yes | int | Maximum bandwidth for this link, given in bytes/s
490 latency | no | double (default: 0.0) | Latency for this link.
491 sharing_policy | no | \ref sharing_policy_shared "SHARED"\|\ref sharing_policy_fatpipe "FATPIPE"\|\ref sharing_policy_fullduplex "FULLDUPLEX" (default: SHARED) | Sharing policy for the link.
492 state | no | ON\|OFF (default: ON) | Allows you to to turn this link on or off (working / not working)
493 bandwidth_file | no | string | Allows you to use a file as input for bandwidth.
494 latency_file | no | string | Allows you to use a file as input for latency.
495 state_file | no | string | Allows you to use a file as input for states.
498 #### Possible shortcuts for ``latency`` ####
500 When using the latency attribute, you can specify the latency by using the scientific
501 notation or by using common abbreviations. For instance, the following three tags
505 <link id="LINK1" bandwidth="125000000" latency="5E-6"/>
506 <link id="LINK1" bandwidth="125000000" latency="5us"/>
507 <link id="LINK1" bandwidth="125000000" latency="0.000005"/>
510 Here, the second tag uses "us", meaning "microseconds". Other shortcuts are:
512 Name | Abbreviation | Time (in seconds)
513 ---- | ------------ | -----------------
514 Week | w | 7 * 24 * 60 * 60
515 Day | d | 24 * 60 * 60
519 Millisecond | ms | 0.001 = 10^(-3)
520 Microsecond | us | 0.000001 = 10^(-6)
521 Nanosecond | ns | 0.000000001 = 10^(-9)
522 Picosecond | ps | 0.000000000001 = 10^(-12)
524 #### Sharing policy ####
526 \anchor sharing_policy_shared
527 By default a network link is \b SHARED, i.e., if two or more data flows go
528 through a link, the bandwidth is shared fairly among all data flows. This
529 is similar to the sharing policy TCP uses.
531 \anchor sharing_policy_fatpipe
532 On the other hand, if a link is defined as a \b FATPIPE,
533 each flow going through this link will be provided with the complete bandwidth,
534 i.e., no sharing occurs and the bandwidth is only limiting each flow individually.
535 Please note that this is really on a per-flow basis, not only on a per-host basis!
536 The complete bandwidth provided by this link in this mode
537 is ``number_of_flows*bandwidth``, with at most ``bandwidth`` being available per flow.
539 Using the FATPIPE mode allows to model backbones that won't affect performance
542 \anchor sharing_policy_fullduplex
543 The last mode available is \b FULLDUPLEX. This means that SimGrid will
544 automatically generate two links (one carrying the suffix _UP and the other the
545 suffix _DOWN) for each ``<link>`` tag. This models situations when the direction
546 of traffic is important.
549 Transfers from one side to the other will interact similarly as
550 TCP when ACK returning packets circulate on the other direction. More
551 discussion about it is available in the description of link_ctn description.
553 In other words: The SHARED policy defines a physical limit for the bandwidth.
554 The FATPIPE mode defines a limit for each application,
555 with no upper total limit.
558 Tip: By using the FATPIPE mode, you can model big backbones that
559 won't affect performance (except latency).
564 <link id="SWITCH" bandwidth="125000000" latency="5E-5" sharing_policy="FATPIPE" />
567 #### Expressing dynamism and failures ####
569 Similar to hosts, it is possible to declare links whose state, bandwidth
570 or latency changes over time (see Section \ref pf_host_dynamism for details).
572 In the case of network links, the ``bandwidth`` and ``latency`` attributes are
573 replaced by the ``bandwidth_file`` and ``latency_file`` attributes.
574 The following XML snippet demonstrates how to use this feature in the platform
575 file. The structure of the files "link1.bw" and "link1.lat" is shown below.
578 <link id="LINK1" state_file="link1.fail" bandwidth="80000000" latency=".0001" bandwidth_file="link1.bw" latency_file="link1.lat" />
582 Even if the syntax is the same, the semantic of bandwidth and latency
583 trace files differs from that of host availability files. For bandwidth and
584 latency, the corresponding files do not
585 express availability as a fraction of the available capacity but directly in
586 bytes per seconds for the bandwidth and in seconds for the latency. This is
587 because most tools allowing to capture traces on real platforms (such as NWS)
588 express their results this way.
590 ##### Example of "link1.bw" file #####
598 In this example, the bandwidth changes repeatedly, with all changes
599 being repeated every 12 seconds.
601 At the beginning of the the simulation, the link's bandwidth is 80,000,000
602 B/s (i.e., 80 Mb/s); this value was defined in the XML snippet above.
603 After four seconds, it drops to 40 Mb/s (line 2), and climbs
604 back to 60 Mb/s after another 4 seconds (line 3). The value does not change any
605 more until the end of the period, that is, after 12 seconds have been simulated).
606 At this point, periodicity kicks in and this behavior is repeated: Seconds
607 12-16 will experience 80 Mb/s, 16-20 40 Mb/s etc.).
609 ##### Example of "link1.lat" file #####
618 In this example, the latency varies with a period of 5 seconds.
619 In the xml snippet above, the latency is initialized to be 0.0001s (100µs). This
620 value will be kept during the first second, since the latency_file contains
621 changes to this value at second one, two and three.
622 At second one, the value will be 0.001, i.e., 1ms. One second later it will
623 be adjusted to 0.01 (or 10ms) and one second later it will be set again to 1ms. The
624 value will not change until second 5, when the periodicity defined in line 1
625 kicks in. It then loops back, starting at 100µs (the initial value) for one second.
628 #### The ``<prop/>`` tag ####
630 Similar to ``\<host\>``, the link may also contain the ``<prop/>`` tag; see the host
631 documentation (Section \ref pf_host) for an example.
636 \subsubsection pf_backbone <backbone/>
639 This tag is <b>only available</b> when the containing AS uses the "Cluster" mode!
641 Using this tag, you can designate an already existing link to be a backbone.
643 Attribute name | Mandatory | Values | Description
644 --------------- | --------- | ------ | -----------
645 id | yes | string | Name of the link that is supposed to act as a backbone.
647 \subsection pf_storage Storage
650 This is a prototype version that should evolve quickly, this
651 is just some doc valuable only at the time of writing this doc
652 This section describes the storage management under SimGrid ; nowadays
653 it's only usable with MSG. It relies basically on linux-like concepts.
654 You also may want to have a look to its corresponding section in \ref
655 msg_file_management ; functions access are organized as a POSIX-like
658 \subsubsection pf_sto_conc Storage Main concepts
659 Basically there are 3 different entities available in SimGrid that
660 can be used to model storages:
662 Entity name | Description
663 --------------- | -----------
664 \ref pf_storage_entity_storage_type "storage_type" | Defines a template for a particular kind of storage (such as a hard-drive) and specifies important features of the storage, such as capacity, performance (read/write), content, ... Different models of hard-drives use different storage_types (because the difference between an SSD and an HDD does matter), as they differ in some specifications (e.g., different size or read/write performance).
665 \ref pf_storage_entity_storage "storage" | Defines an actual instance of a storage (disk, RAM, ...); uses a ``storage_type`` template (see line above) so that you don't need to re-specify the same details over and over again.
666 \ref pf_storage_entity_mount "mount" | Must be wrapped by a \ref pf_host tag; declares which storage(s) this host has mounted and where (i.e., the mountpoint).
669 \anchor pf_storage_content_file
670 ### %Storage Content File ###
672 In order to assess exactly how much time is spent reading from the storage,
673 SimGrid needs to know what is stored in the storage device (identified by distinct (file-)name, like in a file system)
674 and what size this content has.
677 The content file is never changed by the simulation; it is parsed once
678 per simulation and kept in memory afterwards. When the content of the
679 storage changes, only the internal SimGrid data structures change.
681 \anchor pf_storage_content_file_structure
682 #### Structure of a %Storage File ####
684 Here is an excerpt from two storage file; if you want to see the whole file, check
685 the file ``examples/platforms/content/storage_content.txt`` that comes with the
688 SimGrid essentially supports two different formats: UNIX-style filepaths should
689 follow the well known format:
692 /lib/libsimgrid.so.3.6.2 12710497
696 /bin/simgrid_update_xml 5018
697 /bin/graphicator 66986
698 /bin/simgrid-colorizer 2993
703 Windows filepaths, unsurprisingly, use the windows style:
706 \Windows\avastSS.scr 41664
707 \Windows\bfsvc.exe 75264
708 \Windows\bootstat.dat 67584
709 \Windows\CoreSingleLanguage.xml 31497
711 \Windows\dchcfg64.exe 335464
712 \Windows\dcmdev64.exe 93288
716 The different file formats come at a cost; in version 3.12 (and most likely
717 in later versions, too), copying files from windows-style storages to unix-style
718 storages (and vice versa) is not supported.
720 \anchor pf_storage_content_file_create
721 #### Generate a %Storage Content File ####
723 If you want to generate a storage content file based on your own filesystem (or at least a filesystem you have access to),
724 try running this command (works only on unix systems):
727 find /path/you/want -type f -exec ls -l {} \; 2>/dev/null > ./content.txt
730 \subsubsection pf_storage_entities The Storage Entities
732 These are the entities that you can use in your platform files to include
733 storage in your model. See also the list of our \ref pf_storage_example_files "example files";
734 these might also help you to get started.
736 \anchor pf_storage_entity_storage_type
737 #### \<storage_type\> ####
739 Attribute name | Mandatory | Values | Description
740 --------------- | --------- | ------ | -----------
741 id | yes | string | Identifier of this storage_type; used when referring to it
742 model | yes | string | For reasons of future backwards compatibility only; specifies the name of the model for the storage that should be used
743 size | yes | string | Specifies the amount of available storage space; you can specify storage like "500GiB" or "500GB" if you want. (TODO add a link to all the available abbreviations)
744 content | yes | string | Path to a \ref pf_storage_content_file "Storage Content File" on your system. This file must exist.
745 content_type | no | ("txt_unix"\|"txt_win") | Determines which kind of filesystem you're using; make sure the filenames (stored in that file, see \ref pf_storage_content_file_structure "Storage Content File Structure"!)
747 This tag must contain some predefined model properties, specified via the <model_prop> tag. Here is a list,
748 see below for an example:
750 Property id | Mandatory | Values | Description
751 --------------- | --------- | ------ | -----------
752 Bwrite | yes | string | Bandwidth for write access; in B/s (but you can also specify e.g. "30MBps")
753 Bread | yes | string | Bandwidth for read access; in B/s (but you can also specify e.g. "30MBps")
754 Bconnexion | yes | string | Throughput (of the storage connector) in B/s.
757 A storage_type can also contain the <b><prop></b> tag. The <prop> tag allows you
758 to associate additional information to this <storage_type> and follows the
759 attribute/value schema; see the example below. You may want to use it to give information to
760 the tool you use for rendering your simulation, for example.
762 Here is a complete example for the ``storage_type`` tag:
764 <storage_type id="single_HDD" model="linear_no_lat" size="4000" content_type="txt_unix">
765 <model_prop id="Bwrite" value="30MBps" />
766 <model_prop id="Bread" value="100MBps" />
767 <model_prop id="Bconnection" value="150MBps" />
768 <prop id="Brand" value="Western Digital" />
772 \anchor pf_storage_entity_storage
773 #### <storage> ####
775 ``storage`` attributes:
777 Attribute name | Mandatory | Values | Description
778 -------------- | --------- | ------ | -----------
779 id | yes | string | Identifier of this ``storage``; used when referring to it
780 typeId | yes | string | Here you need to refer to an already existing \ref pf_storage_entity_storage_type "\<storage_type\>"; the storage entity defined by this tag will then inherit the properties defined there.
781 attach | yes | string | Name of a host (see Section \ref pf_host) that this storage is <i>physically</i> attached to (e.g., a harddrive in a computer)
782 content | no | string | When specified, overwrites the content attribute of \ref pf_storage_entity_storage_type "\<storage_type\>"
783 content_type | no | string | When specified, overwrites the content_type attribute of \ref pf_storage_entity_storage_type "\<storage_type\>"
785 Here are two examples:
788 <storage id="Disk1" typeId="single_HDD" attach="bob" />
790 <storage id="Disk2" typeId="single_SSD"
791 content="content/win_storage_content.txt"
792 content_type="txt_windows" attach="alice" />
795 The first example is straightforward: A disk is defined and called "Disk1"; it is
796 of type "single_HDD" (shown as an example of \ref pf_storage_entity_storage_type "\<storage_type\>" above) and attached
797 to a host called "bob" (the definition of this host is omitted here).
799 The second storage is called "Disk2", is still of the same type as Disk1 but
800 now specifies a new content file (so the content will be different from Disk1)
801 and the filesystem uses the windows style; finally, it is attached to a second host,
802 called alice (which is again not defined here).
804 \anchor pf_storage_entity_mount
805 #### <mount> ####
808 | Attribute name | Mandatory | Values | Description |
809 | ---------------- | ----------- | -------- | ------------- |
810 | id | yes | string | Refers to a \ref pf_storage_entity_storage "<storage>" entity that will be mounted on that computer |
811 | name | yes | string | Path/location to/of the logical reference (mount point) of this disk
813 This tag must be enclosed by a \ref pf_host tag. It then specifies where the mountpoint of a given storage device (defined by the ``id`` attribute)
814 is; this location is specified by the ``name`` attribute.
816 Here is a simple example, taken from the file ``examples/platform/storage.xml``:
819 <storage_type id="single_SSD" model="linear_no_lat" size="500GiB">
820 <model_prop id="Bwrite" value="60MBps" />
821 <model_prop id="Bread" value="200MBps" />
822 <model_prop id="Bconnection" value="220MBps" />
825 <storage id="Disk2" typeId="single_SSD"
826 content="content/win_storage_content.txt"
827 content_type="txt_windows" attach="alice" />
828 <storage id="Disk4" typeId="single_SSD"
829 content="content/small_content.txt"
830 content_type="txt_unix" attach="denise"/>
832 <host id="alice" power="1Gf">
833 <mount storageId="Disk2" name="c:"/>
836 <host id="denise" power="1Gf">
837 <mount storageId="Disk2" name="c:"/>
838 <mount storageId="Disk4" name="/home"/>
842 This example is quite interesting, as the same device, called "Disk2", is mounted by
843 two hosts at the same time! Note, however, that the host called ``alice`` is actually
844 attached to this storage, as can be seen in the \ref pf_storage_entity_storage "<storage>"
845 tag. This means that ``denise`` must access this storage via network, but SimGrid automatically takes
846 care of that for you.
848 Furthermore, this example shows that ``denise`` has mounted two storages with different
849 filesystem types (unix and windows). In general, a host can mount as many storage devices as
853 Again, the difference between ``attach`` and ``mount`` is simply that
854 an attached storage is always physically inside (or connected to) that machine;
855 for instance, a USB stick is attached to one and only one machine (where it's plugged-in)
856 but it can only be mounted on others, as mounted storage can also be a remote location.
858 \anchor pf_storage_entity_mstorage
859 #### <mstorage> ####
861 This is currently unused.
863 <b>mstorage</b> attributes :
864 \li <b>typeId (mandatory)</b>: the id of the <b>storage</b> that must
865 be mounted on that computer.
866 \li <b>name (mandatory)</b>: the name that will be the logical
867 reference to this disk (the mount point).
869 \subsubsection pf_storage_example_files Example files
871 Several examples were already discussed above; if you're interested in full examples,
872 check the the following platforms:
874 1. ``examples/platforms/storage.xml``
875 2. ``examples/platforms/remote_io.xml``
877 If you're looking for some examplary C code, you may find the source code
878 available in the directory ``examples/msg/io/`` useful.
880 \subsubsection pf_storage_examples_modelling Modelling different situations
882 The storage functionality of SimGrid is type-agnostic, that is, the implementation
883 does not presume any type of storagei, such as HDDs/SSDs, RAM,
884 CD/DVD devices, USB sticks etc.
886 This allows the user to apply the simulator for a wide variety of scenarios; one
887 common scenario would be the access of remote RAM.
889 #### Modelling the access of remote RAM ####
891 How can this be achieved in SimGrid? Let's assume we have a setup where three hosts
892 (HostA, HostB, HostC) need to access remote RAM:
902 An easy way to model this scenario is to setup and define the RAM via the
903 \ref pf_storage_entity_storage "storage" and \ref pf_storage_entity_storage_type "storage type"
904 entities and attach it to a remote dummy host; then, every host can have their own links
905 to this host (modelling for instance certain scenarios, such as PCIe ...)
910 RAM - Dummy -- Host B
915 Now, if read from this storage, the host that mounts this storage
916 communicates to the dummy host which reads from RAM and
917 sends the information back.
920 \section pf_routing Routing
922 To achieve high performance, the routing tables used within SimGrid are
923 static. This means that routing between two nodes is calculated once
924 and will not change during execution. The SimGrid team chose to use this
925 approach as it is rare to have a real deficiency of a resource;
926 most of the time, a communication fails because the links experience too much
927 congestion and hence, your connection stops before the timeout or
928 because the computer designated to be the destination of that message
931 We also chose to use shortest paths algorithms in order to emulate
932 routing. Doing so is consistent with the reality: RIP, OSPF, BGP are
933 all calculating shortest paths. They have some convergence time, but
934 at the end, so when the platform is stable (and this should be the
935 moment you want to simulate something using SimGrid) your packets will
936 follow the shortest paths.
938 \subsection pf_rm Routing models
940 Within each AS, you have to define a routing model to use. You have
941 basically 3 main kind of routing models :
943 \li Shortest-path based models: you let SimGrid calculates shortest
944 paths and manage it. Behaves more or less as most real life
946 \li Manually-entered route models: you'll have to define all routes
947 manually by yourself into the platform description file.
948 Consistent with some manually managed real life routing.
949 \li Simple/fast models: those models offers fast, low memory routing
950 algorithms. You should consider to use it if you can make some
951 assumptions about your AS. Routing in this case is more or less
954 \subsubsection pf_raf The router affair
956 Expressing routers becomes mandatory when using shortest-path based
957 models or when using ns-3 or the bindings to the GTNetS packet-level
958 simulator instead of the native analytical network model implemented
961 For graph-based shortest path algorithms, routers are mandatory,
962 because both algorithms need a graph, and so we need to have source
963 and destination for each edge.
965 Routers are naturally an important concept in GTNetS or ns-3 since the
966 way they run the packet routing algorithms is actually simulated.
967 Instead, the SimGrid’s analytical models aggregate the routing time
968 with the transfer time. Rebuilding a graph representation only from
969 the route information turns to be a very difficult task, because of
970 the missing information about how routes intersect. That is why we
971 introduced a \<router\> tag, which is simply used to express these
972 intersection points. The only attribute accepted by this tag an id. It
973 is important to understand that the \<router\> tag is only used to
974 provide topological information.
976 To express those topological information, some <b>route</b> have to be
977 defined saying which link is between which routers. Description or the
978 route syntax is given below, as well as example for the different
981 \subsubsection pf_rm_sh Shortest-path based models
983 Here is the complete list of such models, that computes routes using
984 classic shortest-paths algorithms. How to choose the best suited
985 algorithm is discussed later in the section devoted to it.
987 \li <b>Floyd</b>: Floyd routing data. Pre-calculates all routes once.
988 \li <b>Dijkstra</b>: Dijkstra routing data ,calculating routes when
990 \li <b>DijkstraCache</b>: Dijkstra routing data. Handle some cache for
991 already calculated routes.
993 All those shortest-path models are instanciated the same way. Here are
998 <AS id="AS0" routing="Floyd">
1000 <cluster id="my_cluster_1" prefix="c-" suffix=""
1001 radical="0-1" power="1000000000" bw="125000000" lat="5E-5"
1002 router_id="router1"/>
1004 <AS id="AS1" routing="none">
1005 <host id="host1" power="1000000000"/>
1008 <link id="link1" bandwidth="100000" latency="0.01"/>
1010 <ASroute src="my_cluster_1" dst="AS1"
1013 <link_ctn id="link1"/>
1019 ASroute given at the end gives a topological information: link1 is
1020 between router1 and host1.
1024 <AS id="AS_2" routing="Dijsktra">
1025 <host id="AS_2_host1" power="1000000000"/>
1026 <host id="AS_2_host2" power="1000000000"/>
1027 <host id="AS_2_host3" power="1000000000"/>
1028 <link id="AS_2_link1" bandwidth="1250000000" latency="5E-4"/>
1029 <link id="AS_2_link2" bandwidth="1250000000" latency="5E-4"/>
1030 <link id="AS_2_link3" bandwidth="1250000000" latency="5E-4"/>
1031 <link id="AS_2_link4" bandwidth="1250000000" latency="5E-4"/>
1032 <router id="central_router"/>
1033 <router id="AS_2_gateway"/>
1034 <!-- routes providing topological information -->
1035 <route src="central_router" dst="AS_2_host1"><link_ctn id="AS_2_link1"/></route>
1036 <route src="central_router" dst="AS_2_host2"><link_ctn id="AS_2_link2"/></route>
1037 <route src="central_router" dst="AS_2_host3"><link_ctn id="AS_2_link3"/></route>
1038 <route src="central_router" dst="AS_2_gateway"><link_ctn id="AS_2_link4"/></route>
1042 DijsktraCache example :
1044 <AS id="AS_2" routing="DijsktraCache">
1045 <host id="AS_2_host1" power="1000000000"/>
1047 (platform unchanged compared to upper example)
1050 \subsubsection pf_rm_me Manually-entered route models
1052 \li <b>Full</b>: You have to enter all necessary routes manually
1056 <AS id="AS0" routing="Full">
1057 <host id="host1" power="1000000000"/>
1058 <host id="host2" power="1000000000"/>
1059 <link id="link1" bandwidth="125000000" latency="0.000100"/>
1060 <route src="host1" dst="host2"><link_ctn id="link1"/></route>
1064 \subsubsection pf_rm_sf Simple/fast models
1066 \li <b>none</b>: No routing (Unless you know what you are doing, avoid
1067 using this mode in combination with a non Constant network model).
1070 <AS id="exitAS" routing="none">
1071 <router id="exit_gateway"/>
1074 \li <b>Vivaldi</b>: Vivaldi routing, so when you want to use
1075 coordinates. See the corresponding section P2P below for details.
1076 \li <b>Cluster</b>: Cluster routing, specific to cluster tag, should
1077 not be used, except internally.
1079 \subsection ps_dec Defining routes
1081 The principle of route definition is the same for the 4 available tags
1082 for doing it. Those for tags are:
1084 \li <b>route</b>: to define route between host/router
1085 \li <b>ASroute</b>: to define route between AS
1086 \li <b>bypassRoute</b>: to bypass normal routes as calculated by the
1087 network model between host/router
1088 \li <b>bypassASroute</b>: same as bypassRoute, but for AS
1090 Basically all those tags will contain an (ordered) list of references
1091 to link that compose the route you want to define.
1093 Consider the example below:
1096 <route src="Alice" dst="Bob">
1097 <link_ctn id="link1"/>
1098 <link_ctn id="link2"/>
1099 <link_ctn id="link3"/>
1103 The route here from host Alice to Bob will be first link1, then link2,
1104 and finally link3. What about the reverse route ? <b>route</b> and
1105 <b>ASroute</b> have an optional attribute <b>symmetrical</b>, that can
1106 be either YES or NO. YES means that the reverse route is the same
1107 route in the inverse order, and is set to YES by default. Note that
1108 this is not the case for bypass*Route, as it is more probable that you
1109 want to bypass only one default route.
1111 For an ASroute, things are just slightly more complicated, as you have
1112 to give the id of the gateway which is inside the AS you're talking
1113 about you want to access ... So it looks like this :
1117 <ASroute src="AS1" dst="AS2"
1118 gw_src="router1" gw_dst="router2">
1119 <link_ctn id="link1"/>
1123 gw == gateway, so when any message are trying to go from AS1 to AS2,
1124 it means that it must pass through router1 to get out of the AS, then
1125 pass through link1, and get into AS2 by being received by router2.
1126 router1 must belong to AS1 and router2 must belong to AS2.
1128 \subsubsection pf_linkctn link_ctn
1130 a <b>link_ctn</b> is the tag that is used in order to reference a
1131 <b>link</b> in a route. Its id is the link id it refers to.
1133 <b>link_ctn</b> attributes :
1134 \li <b>id (mandatory)</b>: Id of the link this tag refers to
1135 \li <b>direction</b>: if the link referenced by <b>id</b> has been
1136 declared as FULLDUPLEX, this is used to indicate in which
1137 direction the route you're defining is going through this link.
1138 Possible values "UP" or "DOWN".
1140 \subsubsection pf_asro ASroute
1142 ASroute tag purpose is to let people write manually their routes
1143 between AS. It's useful when you're in Full model.
1145 <b>ASroute</b> attributes :
1146 \li <b>src (mandatory)</b>: the source AS id.
1147 \li <b>dst (mandatory)</b>: the destination AS id.
1148 \li <b>gw_src (mandatory)</b>: the gateway to be used within the AS.
1149 Can be any <b>host</b> or \b router defined into the \b src AS or
1150 into one of the AS it includes.
1151 \li <b>gw_dst (mandatory)</b>: the gateway to be used within the AS.
1152 Can be any <b>host</b> or \b router defined into the \b dst AS or
1153 into one of the AS it includes.
1154 \li <b>symmetrical</b>: if the route is symmetric, the reverse route
1155 will be the opposite of the one defined. Can be either YES or NO,
1158 <b>Example of ASroute with Full</b>
1160 <AS id="AS0" routing="Full">
1161 <cluster id="my_cluster_1" prefix="c-" suffix=".me"
1162 radical="0-149" power="1000000000" bw="125000000" lat="5E-5"
1163 bb_bw="2250000000" bb_lat="5E-4"/>
1165 <cluster id="my_cluster_2" prefix="c-" suffix=".me"
1166 radical="150-299" power="1000000000" bw="125000000" lat="5E-5"
1167 bb_bw="2250000000" bb_lat="5E-4"/>
1169 <link id="backbone" bandwidth="1250000000" latency="5E-4"/>
1171 <ASroute src="my_cluster_1" dst="my_cluster_2"
1172 gw_src="c-my_cluster_1_router.me"
1173 gw_dst="c-my_cluster_2_router.me">
1174 <link_ctn id="backbone"/>
1176 <ASroute src="my_cluster_2" dst="my_cluster_1"
1177 gw_src="c-my_cluster_2_router.me"
1178 gw_dst="c-my_cluster_1_router.me">
1179 <link_ctn id="backbone"/>
1184 \subsubsection pf_ro route
1185 The principle is the same as ASroute : <b>route</b> contains list of
1186 links that are in the path between src and dst, except that it is for
1187 routes between a src that can be either <b>host</b> or \b router and a
1188 dst that can be either <b>host</b> or \b router. Useful for Full
1189 as well as for the shortest-paths based models, where you
1190 have to give topological information.
1193 <b>route</b> attributes :
1194 \li <b>src (mandatory)</b>: the source id.
1195 \li <b>dst (mandatory)</b>: the destination id.
1196 \li <b>symmetrical</b>: if the route is symmetric, the reverse route
1197 will be the opposite of the one defined. Can be either YES or NO,
1200 <b>route example in Full</b>
1202 <route src="Tremblay" dst="Bourassa">
1203 <link_ctn id="4"/><link_ctn id="3"/><link_ctn id="2"/><link_ctn id="0"/><link_ctn id="1"/><link_ctn id="6"/><link_ctn id="7"/>
1207 <b>route example in a shortest-path model</b>
1209 <route src="Tremblay" dst="Bourassa">
1213 Note that when using route to give topological information, you have
1214 to give routes with one link only in it, as SimGrid needs to know
1215 which host are at the end of the link.
1217 \subsubsection pf_byASro bypassASroute
1219 <b>Note : bypassASroute and bypassRoute are under rewriting to perform
1220 better ; so you may not use it yet</b> As said before, once you choose
1221 a model, it (if so) calculates routes for you. But maybe you want to
1222 define some of your routes, which will be specific. You may also want
1223 to bypass some routes defined in lower level AS at an upper stage :
1224 <b>bypassASroute</b> is the tag you're looking for. It allows to
1225 bypass routes defined between already defined between AS (if you want
1226 to bypass route for a specific host, you should just use byPassRoute).
1227 The principle is the same as ASroute : <b>bypassASroute</b> contains
1228 list of links that are in the path between src and dst.
1230 <b>bypassASroute</b> attributes :
1231 \li <b>src (mandatory)</b>: the source AS id.
1232 \li <b>dst (mandatory)</b>: the destination AS id.
1233 \li <b>gw_src (mandatory)</b>: the gateway to be used within the AS.
1234 Can be any <b>host</b> or \b router defined into the \b src AS or
1235 into one of the AS it includes.
1236 \li <b>gw_dst (mandatory)</b>: the gateway to be used within the AS.
1237 Can be any <b>host</b> or \b router defined into the \b dst AS or
1238 into one of the AS it includes.
1239 \li <b>symmetrical</b>: if the route is symmetric, the reverse route
1240 will be the opposite of the one defined. Can be either YES or NO,
1243 <b>bypassASroute Example</b>
1245 <bypassASRoute src="my_cluster_1" dst="my_cluster_2"
1246 gw_src="my_cluster_1_router"
1247 gw_dst="my_cluster_2_router">
1248 <link_ctn id="link_tmp"/>
1252 \subsubsection pf_byro bypassRoute
1253 <b>Note : bypassASRoute and bypassRoute are under rewriting to perform
1254 better ; so you may not use it yet</b> As said before, once you choose
1255 a model, it (if so) calculates routes for you. But maybe you want to
1256 define some of your routes, which will be specific. You may also want
1257 to bypass some routes defined in lower level AS at an upper stage :
1258 <b>bypassRoute</b> is the tag you're looking for. It allows to bypass
1259 routes defined between <b>host/router</b>. The principle is the same
1260 as route : <b>bypassRoute</b> contains list of links references of
1261 links that are in the path between src and dst.
1263 <b>bypassRoute</b> attributes :
1264 \li <b>src (mandatory)</b>: the source AS id.
1265 \li <b>dst (mandatory)</b>: the destination AS id.
1266 \li <b>symmetrical</b>: if the route is symmetric, the reverse route
1267 will be the opposite of the one defined. Can be either YES or NO,
1270 <b>bypassRoute Example</b>
1272 <bypassRoute src="host_1" dst="host_2">
1273 <link_ctn id="link_tmp"/>
1278 \subsection pb_baroex Basic Routing Example
1280 Let's say you have an AS named AS_Big that contains two other AS, AS_1
1281 and AS_2. If you want to make a host (h1) from AS_1 with another one
1282 (h2) from AS_2 then you'll have to proceed as follows:
1283 \li First, you have to ensure that a route is defined from h1 to the
1284 AS_1's exit gateway and from h2 to AS_2's exit gateway.
1285 \li Then, you'll have to define a route between AS_1 to AS_2. As those
1286 AS are both resources belonging to AS_Big, then it has to be done
1287 at AS_big level. To define such a route, you have to give the
1288 source AS (AS_1), the destination AS (AS_2), and their respective
1289 gateway (as the route is effectively defined between those two
1290 entry/exit points). Elements of this route can only be elements
1291 belonging to AS_Big, so links and routers in this route should be
1292 defined inside AS_Big. If you choose some shortest-path model,
1293 this route will be computed automatically.
1295 As said before, there are mainly 2 tags for routing :
1296 \li <b>ASroute</b>: to define routes between two <b>AS</b>
1297 \li <b>route</b>: to define routes between two <b>host/router</b>
1299 As we are dealing with routes between AS, it means that those we'll
1300 have some definition at AS_Big level. Let consider AS_1 contains 1
1301 host, 1 link and one router and AS_2 3 hosts, 4 links and one router.
1302 There will be a central router, and a cross-like topology. At the end
1303 of the crosses arms, you'll find the 3 hosts and the router that will
1304 act as a gateway. We have to define routes inside those two AS. Let
1305 say that AS_1 contains full routes, and AS_2 contains some Floyd
1306 routing (as we don't want to bother with defining all routes). As
1307 we're using some shortest path algorithms to route into AS_2, we'll
1308 then have to define some <b>route</b> to gives some topological
1309 information to SimGrid. Here is a file doing it all :
1312 <AS id="AS_Big" routing="Dijsktra">
1313 <AS id="AS_1" routing="Full">
1314 <host id="AS_1_host1" power="1000000000"/>
1315 <link id="AS_1_link" bandwidth="1250000000" latency="5E-4"/>
1316 <router id="AS_1_gateway"/>
1317 <route src="AS_1_host1" dst="AS_1_gateway">
1318 <link_ctn id="AS_1_link"/>
1321 <AS id="AS_2" routing="Floyd">
1322 <host id="AS_2_host1" power="1000000000"/>
1323 <host id="AS_2_host2" power="1000000000"/>
1324 <host id="AS_2_host3" power="1000000000"/>
1325 <link id="AS_2_link1" bandwidth="1250000000" latency="5E-4"/>
1326 <link id="AS_2_link2" bandwidth="1250000000" latency="5E-4"/>
1327 <link id="AS_2_link3" bandwidth="1250000000" latency="5E-4"/>
1328 <link id="AS_2_link4" bandwidth="1250000000" latency="5E-4"/>
1329 <router id="central_router"/>
1330 <router id="AS_2_gateway"/>
1331 <!-- routes providing topological information -->
1332 <route src="central_router" dst="AS_2_host1"><link_ctn id="AS_2_link1"/></route>
1333 <route src="central_router" dst="AS_2_host2"><link_ctn id="AS_2_link2"/></route>
1334 <route src="central_router" dst="AS_2_host3"><link_ctn id="AS_2_link3"/></route>
1335 <route src="central_router" dst="AS_2_gateway"><link_ctn id="AS_2_link4"/></route>
1337 <link id="backbone" bandwidth="1250000000" latency="5E-4"/>
1339 <ASroute src="AS_1" dst="AS_2"
1340 gw_src="AS_1_gateway"
1341 gw_dst="AS_2_gateway">
1342 <link_ctn id="backbone"/>
1347 \section pf_other_tags Tags not (directly) describing the platform
1349 There are 3 tags, that you can use inside a \<platform\> tag that are
1350 not describing the platform:
1351 \li random: it allows you to define random generators you want to use
1352 for your simulation.
1353 \li config: it allows you to pass some configuration stuff like, for
1354 example, the network model and so on. It follows the
1355 \li include: simply allows you to include another file into the
1358 \subsection pf_conf config
1359 <b>config</b> attributes :
1360 \li <b>id (mandatory)</b>: the identifier of the config to be used
1361 when referring to it.
1364 <b>config</b> tag only purpose is to include <b>prop</b> tags. Valid
1365 id are basically the same as the list of possible parameters you can
1366 use by command line, except that "/" are used for namespace
1367 definition. See the \ref options config and options page for more
1371 <b>config example</b>
1373 <?xml version='1.0'?>
1374 <!DOCTYPE platform SYSTEM "http://simgrid.gforge.inria.fr/simgrid.dtd">
1375 <platform version="3">
1376 <config id="General">
1377 <prop id="maxmin/precision" value="0.000010"></prop>
1378 <prop id="cpu/optim" value="TI"></prop>
1379 <prop id="host/model" value="compound"></prop>
1380 <prop id="network/model" value="SMPI"></prop>
1381 <prop id="path" value="~/"></prop>
1382 <prop id="smpi/bw_factor" value="65472:0.940694;15424:0.697866;9376:0.58729"></prop>
1385 <AS id="AS0" routing="Full">
1390 \subsection pf_rand random
1391 Not yet in use, and possibly subject to huge modifications.
1393 \subsection pf_incl include
1394 <b>include</b> tag allows to import into a file platform parts located
1395 in another file. This is done with the intention to help people
1396 combine their different AS and provide new platforms. Those files
1397 should contains XML part that contains either
1398 <b>include,cluster,peer,AS,trace,trace_connect</b> tags.
1400 <b>include</b> attributes :
1401 \li <b>file (mandatory)</b>: filename of the file to include. Possible
1402 values: absolute or relative path, syntax similar to the one in
1405 <b>Note</b>: due to some obscure technical reasons, you have to open
1406 and close tag in order to let it work.
1407 <b>include Example</b>
1409 <?xml version='1.0'?>
1410 <!DOCTYPE platform SYSTEM "http://simgrid.gforge.inria.fr/simgrid.dtd">
1411 <platform version="3">
1412 <AS id="main" routing="Full">
1413 <include file="clusterA.xml"></include>
1414 <include file="clusterB.xml"></include>
1419 \subsection pf_tra trace and trace_connect
1420 Both tags are an alternate way to passe availability, state, and so on
1421 files to entity. Instead of referring to the file directly in the host,
1422 link, or cluster tag, you proceed by defining a trace with an id
1423 corresponding to a file, later a host/link/cluster, and finally using
1424 trace_connect you say that the file trace must be used by the entity.
1425 Get it ? Let's have a look at an example :
1428 <AS id="AS0" routing="Full">
1429 <host id="bob" power="1000000000"/>
1431 <trace id="myTrace" file="bob.trace" periodicity="1.0"/>
1432 <trace_connect trace="myTrace" element="bob" kind="POWER"/>
1435 All constraints you have is that <b>trace_connect</b> is after
1436 <b>trace</b> and <b>host</b> definitions.
1439 <b>trace</b> attributes :
1440 \li <b>id (mandatory)</b>: the identifier of the trace to be used when
1442 \li <b>file</b>: filename of the file to include. Possible values :
1443 absolute or relative path, syntax similar to the one in use on
1444 your system. If omitted, the system expects that you provide the
1445 trace values inside the trace tags (see below).
1446 \li <b>trace periodicity (mandatory)</b>: trace periodicity, same
1447 definition as in hosts (see upper for details).
1449 Here is an example of trace when no file name is provided:
1452 <trace id="myTrace" periodicity="1.0">
1459 <b>trace_connect</b> attributes :
1460 \li <b>kind</b>: the type of trace, possible values
1461 <b>HOST_AVAIL|POWER|LINK_AVAIL|BANDWIDTH|LATENCY,</b> default:
1463 \li <b>trace (mandatory)</b>: the identifier of the trace referenced.
1464 \li <b>element (mandatory)</b>: the identifier of the entity referenced.
1468 \section pf_hints Hints and tips, or how to write a platform efficiently
1470 Now you should know at least the syntax and be able to create a
1471 platform by your own. However, after having ourselves wrote some platforms, there
1472 are some best practices you should pay attention to in order to
1473 produce good platform and some choices you can make in order to have
1474 faster simulations. Here's some hints and tips, then.
1476 \subsection pf_as_h AS Hierarchy
1477 The AS design allows SimGrid to go fast, because computing route is
1478 done only for the set of resources defined in this AS. If you're using
1479 only a big AS containing all resource with no AS into it and you're
1480 using Full model, then ... you'll loose all interest into it. On the
1481 other hand, designing a binary tree of AS with, at the lower level,
1482 only one host, then you'll also loose all the good AS hierarchy can
1483 give you. Remind you should always be "reasonable" in your platform
1484 definition when choosing the hierarchy. A good choice if you try to
1485 describe a real life platform is to follow the AS described in
1486 reality, since this kind of trade-off works well for real life
1489 \subsection pf_exit_as Exit AS: why and how
1490 Users that have looked at some of our platforms may have notice a
1491 non-intuitive schema ... Something like that :
1495 <AS id="AS_4" routing="Full">
1496 <AS id="exitAS_4" routing="Full">
1497 <router id="router_4"/>
1499 <cluster id="cl_4_1" prefix="c_4_1-" suffix="" radical="1-20" power="1000000000" bw="125000000" lat="5E-5" bb_bw="2250000000" bb_lat="5E-4"/>
1500 <cluster id="cl_4_2" prefix="c_4_2-" suffix="" radical="1-20" power="1000000000" bw="125000000" lat="5E-5" bb_bw="2250000000" bb_lat="5E-4"/>
1501 <link id="4_1" bandwidth="2250000000" latency="5E-5"/>
1502 <link id="4_2" bandwidth="2250000000" latency="5E-5"/>
1503 <link id="bb_4" bandwidth="2250000000" latency="5E-4"/>
1504 <ASroute src="cl_4_1"
1506 gw_src="c_4_1-cl_4_1_router"
1507 gw_dst="c_4_2-cl_4_2_router"
1509 <link_ctn id="4_1"/>
1510 <link_ctn id="bb_4"/>
1511 <link_ctn id="4_2"/>
1513 <ASroute src="cl_4_1"
1515 gw_src="c_4_1-cl_4_1_router"
1518 <link_ctn id="4_1"/>
1519 <link_ctn id="bb_4"/>
1521 <ASroute src="cl_4_2"
1523 gw_src="c_4_2-cl_4_2_router"
1526 <link_ctn id="4_2"/>
1527 <link_ctn id="bb_4"/>
1532 In the AS_4, you have an exitAS_4 defined, containing only one router,
1533 and routes defined to that AS from all other AS (as cluster is only a
1534 shortcut for an AS, see cluster description for details). If there was
1535 an upper AS, it would define routes to and from AS_4 with the gateway
1536 router_4. It's just because, as we did not allowed (for performances
1537 issues) to have routes from an AS to a single host/router, you have to
1538 enclose your gateway, when you have AS included in your AS, within an
1539 AS to define routes to it.
1541 \subsection pf_P2P_tags P2P or how to use coordinates
1542 SimGrid allows you to use some coordinated-based system, like vivaldi,
1543 to describe a platform. The main concept is that you have some peers
1544 that are located somewhere: this is the function of the
1545 <b>coordinates</b> of the \<peer\> or \<host\> tag. There's nothing
1546 complicated in using it, here is an example of it:
1549 <?xml version='1.0'?>
1550 <!DOCTYPE platform SYSTEM "http://simgrid.gforge.inria.fr/simgrid.dtd">
1551 <platform version="3">
1553 <config id="General">
1554 <prop id="network/coordinates" value="yes"></prop>
1556 <AS id="AS0" routing="Vivaldi">
1557 <host id="100030591" coordinates="25.5 9.4 1.4" power="1500000000.0" />
1558 <host id="100036570" coordinates="-12.7 -9.9 2.1" power="730000000.0" />
1560 <host id="100429957" coordinates="17.5 6.7 18.8" power="830000000.0" />
1565 Coordinates are then used to calculate latency between two hosts by
1566 calculating the euclidean distance between the two hosts coordinates.
1567 The results express the latency in ms.
1569 Note that the previous example defines a routing directly between hosts but it could be also used to define a routing between AS.
1570 That is for example what is commonly done when using peers (see Section \ref pf_peer).
1572 <?xml version='1.0'?>
1573 <!DOCTYPE platform SYSTEM "http://simgrid.gforge.inria.fr/simgrid.dtd">
1574 <platform version="3">
1576 <config id="General">
1577 <prop id="network/coordinates" value="yes"></prop>
1579 <AS id="AS0" routing="Vivaldi">
1580 <peer id="peer-0" coordinates="173.0 96.8 0.1" power="730Mf" bw_in="13.38MBps" bw_out="1.024MBps" lat="500us"/>
1581 <peer id="peer-1" coordinates="247.0 57.3 0.6" power="730Mf" bw_in="13.38MBps" bw_out="1.024MBps" lat="500us" />
1582 <peer id="peer-2" coordinates="243.4 58.8 1.4" power="730Mf" bw_in="13.38MBps" bw_out="1.024MBps" lat="500us" />
1586 In such a case though, we connect the AS created by the <b>peer</b> tag with the Vivaldi routing mechanism.
1587 This means that to route between AS1 and AS2, it will use the coordinates of router_AS1 and router_AS2.
1588 This is currently a convention and we may offer to change this convention in the DTD later if needed.
1589 You may have noted that conveniently, a peer named FOO defines an AS named FOO and a router named router_FOO, which is why it works seamlessly with the <b>peer</b> tag.
1592 \subsection pf_wisely Choosing wisely the routing model to use
1595 Choosing wisely the routing model to use can significantly fasten your
1596 simulation/save your time when writing the platform/save tremendous
1597 disk space. Here is the list of available model and their
1598 characteristics (lookup : time to resolve a route):
1600 \li <b>Full</b>: Full routing data (fast, large memory requirements,
1602 \li <b>Floyd</b>: Floyd routing data (slow initialization, fast
1603 lookup, lesser memory requirements, shortest path routing only).
1604 Calculates all routes at once at the beginning.
1605 \li <b>Dijkstra</b>: Dijkstra routing data (fast initialization, slow
1606 lookup, small memory requirements, shortest path routing only).
1607 Calculates a route when necessary.
1608 \li <b>DijkstraCache</b>: Dijkstra routing data (fast initialization,
1609 fast lookup, small memory requirements, shortest path routing
1610 only). Same as Dijkstra, except it handles a cache for latest used
1612 \li <b>none</b>: No routing (usable with Constant network only).
1613 Defines that there is no routes, so if you try to determine a
1614 route without constant network within this AS, SimGrid will raise
1616 \li <b>Vivaldi</b>: Vivaldi routing, so when you want to use coordinates
1617 \li <b>Cluster</b>: Cluster routing, specific to cluster tag, should
1620 \subsection pf_switch Hey, I want to describe a switch but there is no switch tag !
1622 Actually we did not include switch tag, ok. But when you're trying to
1623 simulate a switch, the only major impact it has when you're using
1624 fluid model (and SimGrid uses fluid model unless you activate GTNetS,
1625 ns-3, or constant network mode) is the impact of the upper limit of
1626 the switch motherboard speed that will eventually be reached if you're
1627 using intensively your switch. So, the switch impact is similar to a
1628 link one. That's why we are used to describe a switch using a link tag
1629 (as a link is not an edge by a hyperedge, you can connect more than 2
1632 \subsection pf_platform_multipath How to express multipath routing in platform files?
1634 It is unfortunately impossible to express the fact that there is more
1635 than one routing path between two given hosts. Let's consider the
1636 following platform file:
1639 <route src="A" dst="B">
1642 <route src="B" dst="C">
1645 <route src="A" dst="C">
1650 Although it is perfectly valid, it does not mean that data traveling
1651 from A to C can either go directly (using link 3) or through B (using
1652 links 1 and 2). It simply means that the routing on the graph is not
1653 trivial, and that data do not following the shortest path in number of
1654 hops on this graph. Another way to say it is that there is no implicit
1655 in these routing descriptions. The system will only use the routes you
1656 declare (such as <route src="A" dst="C"><link_ctn
1657 id="3"/></route>), without trying to build new routes by aggregating
1660 You are also free to declare platform where the routing is not
1661 symmetric. For example, add the following to the previous file:
1664 <route src="C" dst="A">
1670 This makes sure that data from C to A go through B where data from A
1671 to C go directly. Don't worry about realism of such settings since
1672 we've seen ways more weird situation in real settings (in fact, that's
1673 the realism of very regular platforms which is questionable, but
1674 that's another story).
1676 \section pf_flexml_bypassing Bypassing the XML parser with your own C functions
1677 <b>NOTE THAT THIS DOCUMENTATION, WHILE STILL WORKING, IS STRONGLY DEPRECATED</b>
1679 So you want to bypass the XML files parser, uh? Maybe doing some parameter
1680 sweep experiments on your simulations or so? This is possible, and
1681 it's not even really difficult (well. Such a brutal idea could be
1682 harder to implement). Here is how it goes.
1684 For this, you have to first remember that the XML parsing in SimGrid is done
1685 using a tool called FleXML. Given a DTD, this gives a flex-based parser. If
1686 you want to bypass the parser, you need to provide some code mimicking what
1687 it does and replacing it in its interactions with the SURF code. So, let's
1688 have a look at these interactions.
1690 FleXML parser are close to classical SAX parsers. It means that a
1691 well-formed SimGrid platform XML file might result in the following
1694 - start "platform_description" with attribute version="2"
1695 - start "host" with attributes id="host1" power="1.0"
1697 - start "host" with attributes id="host2" power="2.0"
1699 - start "link" with ...
1701 - start "route" with ...
1702 - start "link_ctn" with ...
1705 - end "platform_description"
1707 The communication from the parser to the SURF code uses two means:
1708 Attributes get copied into some global variables, and a surf-provided
1709 function gets called by the parser for each event. For example, the event
1710 - start "host" with attributes id="host1" power="1.0"
1712 let the parser do something roughly equivalent to:
1714 strcpy(A_host_id,"host1");
1719 In SURF, we attach callbacks to the different events by initializing the
1720 pointer functions to some the right surf functions. Since there can be
1721 more than one callback attached to the same event (if more than one
1722 model is in use, for example), they are stored in a dynar. Example in
1725 /* Adding callback functions */
1726 surf_parse_reset_parser();
1727 surfxml_add_callback(STag_surfxml_host_cb_list, &parse_cpu_init);
1728 surfxml_add_callback(STag_surfxml_prop_cb_list, &parse_properties);
1729 surfxml_add_callback(STag_surfxml_link_cb_list, &parse_link_init);
1730 surfxml_add_callback(STag_surfxml_route_cb_list, &parse_route_set_endpoints);
1731 surfxml_add_callback(ETag_surfxml_link_c_ctn_cb_list, &parse_route_elem);
1732 surfxml_add_callback(ETag_surfxml_route_cb_list, &parse_route_set_route);
1734 /* Parse the file */
1735 surf_parse_open(file);
1736 xbt_assert(!surf_parse(), "Parse error in %s", file);
1740 So, to bypass the FleXML parser, you need to write your own version of the
1741 surf_parse function, which should do the following:
1742 - Fill the A_<tag>_<attribute> variables with the wanted values
1743 - Call the corresponding STag_<tag>_fun function to simulate tag start
1744 - Call the corresponding ETag_<tag>_fun function to simulate tag end
1745 - (do the same for the next set of values, and loop)
1747 Then, tell SimGrid that you want to use your own "parser" instead of the stock one:
1749 surf_parse = surf_parse_bypass_environment;
1750 MSG_create_environment(NULL);
1751 surf_parse = surf_parse_bypass_application;
1752 MSG_launch_application(NULL);
1755 A set of macros are provided at the end of
1756 include/surf/surfxml_parse.h to ease the writing of the bypass
1757 functions. An example of this trick is distributed in the file
1758 examples/msg/masterslave/masterslave_bypass.c