Pipeline¶

Pipelines allow to arbitrarly interconnect generators and resources by chaining them. Messages belonging to a pipeline automatically flow from a resource to another without the needs of explicitly defining the hops. Please note that a generator (and their messages) belong only to a single pipeline. If multiple pipelines are needed simultaneously, they need to be merged fisrt. See Branching pipeline for more details.

Single pipeline¶

Let’s start with a simulation where messages are generated by Generator #0 and are served by Resource #0 and Resource #1.

|Generator #0| -> |Resource #0| -> |Resource #1|

The SimPype code would hence be:

import simpype
import random

sim = simpype.Simulation(id = 'single')
gen0 = sim.add_generator(id = 'gen0')
gen0.random['arrival'] = {0: lambda: random.expovariate(1.0)}
res0 = sim.add_resource(id = 'res0')
res0.random['service'] = {0: lambda: random.expovariate(2.0)}
res1 = sim.add_resource(id = 'res1')
res1.random['service'] = {0: lambda: random.expovariate(2.0)}

p0 = sim.add_pipeline(gen0, res0, res1)

sim.run(until = 5)

As it can be noticed in sim.log file, messages generated by Generator #0 automatically flow through Resource #0 and Resource #1 without the needs of explicitly defining the next hop:

timestamp,message,seq_num,resource,event
000000000,gen0,0,res0,pipe.in
000000000,gen0,0,res0,pipe.out
044474460,gen0,0,res0,resource.serve
044474460,gen0,0,res1,pipe.in
044474460,gen0,0,res1,pipe.out
867233916,gen0,1,res0,pipe.in
867233916,gen0,1,res0,pipe.out
099185483,gen0,0,res1,resource.serve
876512438,gen0,1,res0,resource.serve
876512438,gen0,1,res1,pipe.in
876512438,gen0,1,res1,pipe.out
873364054,gen0,1,res1,resource.serve

Overlapping pipelines¶

Noe let’s continue with a simulation scenarios like the following:

|Generator #0| -\                    /-> |Resource #1|
                 )-> |Resource #0| -(
|Generator #1| -/                    \-> |Resource #2|

In this scenario we want to reproduce the following interconnection:

|Generator #0| -> |Resource #0| -> |Resource #1|

|Generator #1| -> |Resource #0| -> |Resource #2|

As it can be noticed, there are two dinstinct paths/pipelines that overlap at Resource #0. However, any messages generated by Generator #0 should end to Resource #1. Similarly, any messages generated by Generator #1 should end to Resource #2.

In this scenario, Resource #0 is hence shared between the two pipelines. The SimPype code would hence be:

import simpype
import random

sim = simpype.Simulation(id = 'overlap')
gen0 = sim.add_generator(id = 'gen0')
gen0.random['arrival'] = {0: lambda: random.expovariate(1.0)}
gen1 = sim.add_generator(id = 'gen1')
gen1.random['arrival'] = {0: lambda: random.expovariate(1.0)}
res0 = sim.add_resource(id = 'res0')
res0.random['service'] = {0: lambda: random.expovariate(4.0)}
res1 = sim.add_resource(id = 'res1')
res1.random['service'] = {0: lambda: random.expovariate(2.0)}
res2 = sim.add_resource(id = 'res2')
res2.random['service'] = {0: lambda: random.expovariate(2.0)}

p0 = sim.add_pipeline(gen0, res0, res1)
p1 = sim.add_pipeline(gen1, res0, res2)

sim.run(until = 2.5)

As it can be noticed in sim.log file, messages generated by Generator #0 automatically flow through Resource #0 and Resource #1 and messages generated by Generator #1 automatically flow through Resource #0 and Resource #2. Moreover, Resource #0 is shared between the two pipelines:

timestamp,message,seq_num,resource,event
000000000,gen0,0,res0,pipe.in
000000000,gen1,0,res0,pipe.in
000000000,gen0,0,res0,pipe.out
372608250,gen0,0,res0,resource.serve
372608250,gen0,0,res1,pipe.in
372608250,gen0,0,res1,pipe.out
372608250,gen1,0,res0,pipe.out
515112655,gen0,1,res0,pipe.in
636849329,gen1,0,res0,resource.serve
636849329,gen1,0,res2,pipe.in
636849329,gen1,0,res2,pipe.out
636849329,gen0,1,res0,pipe.out
653319564,gen0,1,res0,resource.serve
653319564,gen0,1,res1,pipe.in
684766776,gen1,1,res0,pipe.in
684766776,gen1,1,res0,pipe.out
851617505,gen0,0,res1,resource.serve
851617505,gen0,1,res1,pipe.out
921614468,gen1,2,res0,pipe.in
949578262,gen1,1,res0,resource.serve
949578262,gen1,1,res2,pipe.in
949578262,gen1,2,res0,pipe.out
052881475,gen1,2,res0,resource.serve
052881475,gen1,2,res2,pipe.in
079748898,gen0,1,res1,resource.serve
245866822,gen1,3,res0,pipe.in
245866822,gen1,3,res0,pipe.out
352498249,gen1,0,res2,resource.serve
352498249,gen1,1,res2,pipe.out
369990105,gen1,4,res0,pipe.in
384336838,gen1,1,res2,resource.serve
384336838,gen1,2,res2,pipe.out
385217621,gen1,5,res0,pipe.in
418331444,gen1,2,res2,resource.serve
582122574,gen1,3,res0,resource.serve
582122574,gen1,3,res2,pipe.in
582122574,gen1,3,res2,pipe.out
582122574,gen1,4,res0,pipe.out
028251841,gen1,4,res0,resource.serve
028251841,gen1,4,res2,pipe.in
028251841,gen1,5,res0,pipe.out
148959938,gen1,6,res0,pipe.in

Branching pipeline¶

Now let’s continue with a pipeline having a branching point with one generator and three resources:

                                  /-> |Resource #1|
|Generator #0| -> |Resource #0| -(
                                  \-> |Resource #2|

There are two possible options at this stage:

Serve a copy of the same message to both Resource #1 and Resource #2;

Either serve a message to Resource #1 or to Resource #2.

Automatic copy¶

In case of serving a copy of the same message to both Resource #1 and Resource #2, the SimPype code would hence be:

import simpype
import random

sim = simpype.Simulation(id = 'single')
gen0 = sim.add_generator(id = 'gen0')
gen0.random['arrival'] = {0: lambda: random.expovariate(1.0)}
res0 = sim.add_resource(id = 'res0')
res0.random['service'] = {0: lambda: random.expovariate(2.0)}
res1 = sim.add_resource(id = 'res1')
res1.random['service'] = {0: lambda: random.expovariate(2.0)}
res2 = sim.add_resource(id = 'res2')
res2.random['service'] = {0: lambda: random.expovariate(2.0)}

p0 = sim.add_pipeline(gen0, res0, res1)
p1 = sim.add_pipeline(gen0, res0, res2)
pM = sim.merge_pipeline(p0, p1)

sim.run(until = 5)

Please note the use of merge_pipeline(). This function merges multiple pipelines into a single one, thus creating the branching point. Withouth calling the merge_pipeline() function, the only active pipeline would have been p1.

As it can be noticed in sim.log file, messages are automatically copied and served to both Resource #1 and Resource #2 after being served by Resource #0:

timestamp,message,seq_num,resource,event
000000000,gen0,0,res0,pipe.in
000000000,gen0,0,res0,pipe.out
412762064,gen0,0,res0,resource.serve
412762064,gen0,0,res2,pipe.in
412762064,gen0,0,res1,pipe.in
412762064,gen0,0,res2,pipe.out
412762064,gen0,0,res1,pipe.out
631472230,gen0,0,res1,resource.serve
989221320,gen0,0,res2,resource.serve
545794865,gen0,1,res0,pipe.in
545794865,gen0,1,res0,pipe.out
572402316,gen0,1,res0,resource.serve
572402316,gen0,1,res2,pipe.in
572402316,gen0,1,res1,pipe.in
572402316,gen0,1,res2,pipe.out
572402316,gen0,1,res1,pipe.out
602942195,gen0,1,res1,resource.serve
163453623,gen0,2,res0,pipe.in
163453623,gen0,2,res0,pipe.out
222865258,gen0,2,res0,resource.serve
222865258,gen0,2,res2,pipe.in
222865258,gen0,2,res1,pipe.in
222865258,gen0,2,res1,pipe.out
270038314,gen0,1,res2,resource.serve
270038314,gen0,2,res2,pipe.out
360461106,gen0,2,res2,resource.serve
551208266,gen0,2,res1,resource.serve

Next hop selection¶

Please refer to Next in Message section to understand how the next hop of the messages can be dynamically changed.

Miscellaneous¶

add_pipeline() admits both Resource and Pipeline objects as arguments as shown in this examples:

import simpype

sim = simpype.Simulation(id = 'single')
gen0 = sim.add_generator(id = 'gen0')
gen1 = sim.add_generator(id = 'gen1')
res0 = sim.add_resource(id = 'res0')
res1 = sim.add_resource(id = 'res1')
res2 = sim.add_resource(id = 'res2')
res3 = sim.add_resource(id = 'res3')
res4 = sim.add_resource(id = 'res4')
res5 = sim.add_resource(id = 'res5')
res6 = sim.add_resource(id = 'res6')

# Only resources
p0 = sim.add_pipeline(res0, res1, res2)
p1 = sim.add_pipeline(res3, res4, res5)
# Mixed pipeline and resources
p2 = sim.add_pipeline(gen0, p0)
p3 = sim.add_pipeline(gen1, p1)
p4 = sim.add_pipeline(p3, res6)
# Only pipelines
# Equivalent to sim.add_pipeline(res0, res1, res2, res3, res4, res5)
p4 = sim.add_pipeline(p0, p1)

Instead, merge_pipeline() only admits Pipeline objects as arguments.