[Ndn-interest] Multi-path retrieval performance

[Davide Pesavento]

NFD could be the bottleneck.

On Tue, Sep 19, 2017 at 4:59 PM, Klaus Schneider <klaus at cs.arizona.edu> wrote:
> Also, since your delay is so low (0.4ms) the pipeline size (the number of
> packets in flight) doesn't seem to be the bottleneck. You see that you get
> almost the same results, no matter what the pipeline size.
>
> Did you check the CPU utilization of your VMs during the run? It might be at
> 100%.
>
> Best regards,
> Klaus
>
>
>
>
> On 09/19/2017 01:25 PM, Mohammad Alhowaidi wrote:
>>
>> I am using VMs for every consumer/router/producer , the link bandwidth is
>> 10 Gb/s, the delay is 0.4 ms.
>>
>> The consumer is now connected to 3 routers where every routers has the 1G
>> file in its content store ( I increased the cs size so it can fit 1G file).
>>
>> For the forwarding strategy I am using something similar to this one (with
>> slightly changes)
>>
>> https://github.com/schneiderklaus/ndnSIM/tree/master/examples/ndn-load-balancer
>>
>> I used the following command to run the experiments:
>> ndncatchunks -d iterative /ndn/foo       ; where the file name is foo
>>
>> The time I am getting for the experiments
>> retrieving from 1 router and *10* pipeline interest:  1min,7sec
>> retrieving from 2 routers and *10* pipeline interest:  1min,1sec
>> retrieving from 3 routers and *10* pipeline interest:  1min,2sec
>>
>> retrieving from 1 router and *20* pipeline interest:  1min,31sec
>> retrieving from 2 routers and *20* pipeline interest:  1min,1sec
>> retrieving from 3 routers and *20* pipeline interest:  1min,5sec
>>
>> retrieving from 1 router and *30* pipeline interest:  1min,7sec
>> retrieving from 2 routers and *30* pipeline interest:  1min,2sec
>> retrieving from 3 routers and *30* pipeline interest:  1min,5sec
>>
>> if I used aimd I will get the following result:
>> 1 router: 51 second
>> For 2 routers and 3 routers, most of the time it will timeout for some
>> segments but I could success one time to get the value for 2 routers
>> 2 routers: 1m,10 sec
>>
>>
>> Thanks,
>> Mohammad
>>
>> On Mon, Sep 18, 2017 at 4:38 PM, Nick Briggs <nicholas.h.briggs at gmail.com
>> <mailto:nicholas.h.briggs at gmail.com>> wrote:
>>
>>     One other thing -- if the links are over TCP you probably want to
>>     know the net.inet.tcp.sendspace and .recvspace settings (or
>>     equivalent, and possibly other network configuration parameters).
>> If you end up with a producer process being suspended/rescheduled
>>     because it has overrun the TCP output buffer space it can introduce
>>     hiccups into the performance...
>>
>>     -- Nick
>>
>>     > On Sep 18, 2017, at 2:22 PM, Klaus Schneider <klaus at cs.arizona.edu
>> <mailto:klaus at cs.arizona.edu>> wrote:
>>     >
>>      > + What's the exact simulation setup? Topology, link bandwidth,
>>     delays, etc. ?
>>      >
>>      > On 09/17/2017 03:08 PM, Mohammad Alhowaidi wrote:
>>      >> Hello,
>>      >> I was trying to do a load balancing (multi-path retrieval)
>>     experiments to get a fast retrieval for a file.
>>      >> for example I have a router connected to three producers. and I
>>     used ndncatchunk to retrieve a file (1GB) and I started to increase
>>     the interest pipeline value.
>>      >> I found that when the value of interest-pipeline is 10 is the
>>     best when retrieving from one producer (not sure *WHY*? ), then I
>>     tried to use 2 producers with interest-pipeline is 20 and get a
>>     better performance (retrieving chunks was divided evenly between
>>     these two producers).
>>      >> But when I used 3 producers the performance reduced, whatever I
>>     choose the interest-pipeline value. Shouldn't retrieving the file
>>     from 3 producers with 30 interest-pipeline value be better? what
>>     could be the problem?
>>      >> Sorry for the long question!
>>      >> Thanks,
>>      >> Mohammad
>>      >> _______________________________________________
>>      >> Ndn-interest mailing list
>>      >> Ndn-interest at lists.cs.ucla.edu
>>     <mailto:Ndn-interest at lists.cs.ucla.edu>
>>      >> http://www.lists.cs.ucla.edu/mailman/listinfo/ndn-interest
>>     <http://www.lists.cs.ucla.edu/mailman/listinfo/ndn-interest>
>>      > _______________________________________________
>>      > Ndn-interest mailing list
>>      > Ndn-interest at lists.cs.ucla.edu
>>     <mailto:Ndn-interest at lists.cs.ucla.edu>
>>      > http://www.lists.cs.ucla.edu/mailman/listinfo/ndn-interest
>>     <http://www.lists.cs.ucla.edu/mailman/listinfo/ndn-interest>
>>
>>
> _______________________________________________
> Ndn-interest mailing list
> Ndn-interest at lists.cs.ucla.edu
> http://www.lists.cs.ucla.edu/mailman/listinfo/ndn-interest