Skip to main content
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [sumo-user] one-shot.py is so slow
Hello,
1) I don't look at the stackoverflow questions, so I cannot actually tell you how good the response time / answer quality there is. I recommend the mailing list.

2) I'm not quire sure about your setup now. You can load trips direclty into sumo (and will perform initial routing just like one-shot does it.
As far As I understood you used DUAROUTER once to turn the trips into routes and then ran these with sumo.
What I don't understand is why there are ~55k vehicles active in the network after just 2 hours. I would assume that this is a bit too much for Wellingten. I suspect that your network is completely jammed and this is slowing down the simulation.

If you run one-shot.py with default arguments it runs through a list of rerouting-frequencies. The first frequency is -1 which means no rerouting. This could also lead to a very jammed network. However, this would not explain the slow speed at the very start where not that many vehicles have been loaded. Try running the .sumocfg generated by one-shot py with the gui and figure out if something odd is happening.

One thing to look out for is the number of insertion-backlogged vehicles listed in the network parameters (accessible via right-click on the background or with the little green vehicle button at the bottom).

regards,
Jakob


2018-08-29 23:00 GMT+02:00 mohsen hs <mohsenhs82@xxxxxxxxx>:
Hi Jakob,

Thank you for your reply. As the default routing algorithm in one-shot.py is astar, this is the routing which is responsible for finding the best route. 

The total number of edges is 108k.

Using sumo as cmd was a good idea, thanks for this.

So, I used DUAROUTER  and generated the routes for the same set of trips that I fed into oneshot, with this configuration:
  <files>
        <net-file value="map10_limitedWellington.net.xml"/>
        <route-files value="Routev1800000in30min_v1trips.rou.xml"/>  -> used DUAROUTER  for this
         
    </files>
    <process>
        <begin value="0"/>
        <route-steps value="200"/>
        <routing-algorithm value="astar"/>
        <device.rerouting.probability value="1"/>
        <device.rerouting.period value="300"/>   FOR ASSIGNING NEW ROUTES TO VEHICLES  
        <device.rerouting.adaptation-interval value="1"/>
        <device.rerouting.with-taz value="False"/>
        <device.rerouting.explicit value=""/>
        <vehroute-output.last-route value="False"/>
        <vehroute-output.exit-times value="False"/>
        <vehroute-output.sorted value="False"/>
    </process>
    <reports>
        <verbose value="True"/>
        <no-warnings value="True"/>
    </reports>
</configuration>



Generally, this is much faster than one-shot as only after 20-30 min the simulation step is about 7000 ( although I am aware that I am only doing the routing in 5 min intervals and am missing the routing at the departure time as one-shot does these rerouting at these two step (departure and predefined intervals))


~/Desktop/sumofromgit/bin$ sudo sumo -c TestBigdata.sumocfg
Loading configuration... done.
Loading net-file from 'map10_limitedWellington.net.xml'... done (4512ms).
Loading done.
Simulation started with time: 0.00
Step #6986.00 (1186ms ~= 0.84*RT, ~46293.42UPS, vehicles TOT 65099 ACT 54895 B

while after 3 days of running  one-shot, one-shot.log shows step 1036 as the last line of the log:

Step #1035.00 (371223ms ~= 0.00*RT, ~21.80UPS, vehicles TOT 8135 ACT 8094 BUF 
Step #1036.00 (369297ms ~= 0.00*RT, ~22.00UPS, vehicles TOT 8167 ACT 8126 BU

By the way, I was not aware that we can post our questions to stackoverflow, so I post another version of my question there :


In between, which is a better place to ask questions: stackoverflow or here?

Many thanks
Mohsen



On Wednesday, August 29, 2018, 5:45:53 PM GMT+12, Jakob Erdmann <namdre.sumo@xxxxxxxxx> wrote:


Hello,
How many edges does your network have?
If it's large then the routing might be slowing you down. You can test this by simply starting sumo with a fraction of the trips and option --verbose. You will get some output on the command line that tells you how much time the routing took and how many network edges were check while computing the route.
You can speed up routing by selecting another routing algorithm (http://sumo.sourceforge.net/userdoc/Simulation/Routing.html#Routing_Algorithms)
Possibly, the simulation does not continue due to heavy jamming. This is best investigated by looking at the simulation with sumo-gui (see http://sumo.sourceforge.net/userdoc/FAQ.html#The_simulation_has_lots_of_jams.2Fdeadlocks._What_can_I_do.3F)

regards,
Jakob



2018-08-28 20:55 GMT+02:00 mohsen hs <mohsenhs82@xxxxxxxxx>:
Hello there, 

I am running one-shot.py for 140k trips in(uniformly distrusted in the first 30 min of simulation) min and it seems very slow. It is running for two days and vehroutes_300.xml, 300 is -f,  is only 26KB and the log is  45KB and the last line is :
Step #776.00 (357857ms ~= 0.00*RT, ~15.57UPS, vehicles TOT 5584 ACT 5573 BUF 1
Step #777.00 (356625ms ~= 0.00*RT, ~15.95UPS, vehicles TO

and I have not gotten any error in power shell as it is:
PS D:\S3\sumo-win64-0.32.0\sumo- 0.32.0\tools\assign> python one-shot.py -f 300 -n map10_limitedWellington.net. xml -t tri
psmodified_ tripsv1800000in30min_.trips. xml
> Running simulation with update frequency 300
>> Begin time 2018-08-27 12:53:58.461000


Is there any way that I can speed up the process (my cpu and ram are not even 20% utilized) ? or is there any alternative option for me?
As far as I know if I add and run reroute in TraCI, it would be slow too. 


Looking forward to hearing from you.

Many thanks
Mohsen 

______________________________ _________________
sumo-user mailing list
sumo-user@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/ mailman/listinfo/sumo-user



Back to the top