[Mini-NDN] [EXT] Timeouts in MiniNDN
Andre
adcarneiro at inf.ufrgs.br
Wed Apr 21 13:24:34 PDT 2021
Hi Junxiao,
that makes sense, thanks for the help so far.
I looked into the ndnping example but one doubt remains. What if I want
to consume large chunks of data, say 10MB or more. In this case, I will
need to divide that into chunks of 8800 bytes at most, since that is the
maximum NDN package size allowed. However, using the same mechanism as
ndnping or consumer/consumer-with-timer, the program would have to wait
until one 8800 B package is received before requesting the next one
(using face.expressInterest). How can I do that so that I don`t have to
wait a full round-trip-time before consuming the next part?
Best regards,
André
On 20/04/2021 01:07, Junxiao Shi wrote:
> Hi André
>
> The number of faces in an application should be O(1).
> If you have O(N) or O(log N) faces, you are doing it wrong.
>
> Most applications only have one or two faces.
> You can send and receive multiple packets through the same face.
>
> Yours, Junxiao
>
> On Mon, Apr 19, 2021 at 10:47 PM André Dexheimer
> <adcarneiro at inf.ufrgs.br <mailto:adcarneiro at inf.ufrgs.br>> wrote:
>
> *External Email*
>
> Hello Junxiao
>
> Thank you for the quick response.
>
> I have 6 CPU cores and 12GB of RAM allocated to the virtual machine.
>
> The consumer-with-timer was slightly modified to schedule a
> variable number of consumer requests at a variable period.
>
> How would you suggest that I instantiate the faces to reduce
> overhead? Maybe create an array of faces so that I don't have to
> wait for each packet to arrive before I request the next one?
>
> Best regard,
> Andre
>
> On Apr 19, 2021 22:48, Junxiao Shi <shijunxiao at email.arizona.edu
> <mailto:shijunxiao at email.arizona.edu>> wrote:
>
> Hi Andre
>
> How much hardware resources did you allocate to the topology?
> Since you have 8 nodes, you need at least 8 CPU cores and 8GB
> RAM for a smooth operation.
>
> If you are using ndn-cxx/examples/consumer-with-timer.cpp
> unchanged: each invocation of this program sends 2 Interests,
> and then the program exits.
> Forking and joining all those short-lived processes, as well
> as NFD accepting and closing the faces, would cause
> significant overhead.
> You should instead use a long-lived consumer program, such as
> the ndnping client.
>
> Yours, Junxiao
>
> On Mon, Apr 19, 2021 at 9:36 PM Andre via Mini-NDN
> <mini-ndn at lists.cs.ucla.edu
> <mailto:mini-ndn at lists.cs.ucla.edu>> wrote:
>
> *External Email*
>
> Hi
>
>
> I am running a simple experiment generating traffic
> between nodes in the following topology, where all links
> have 10ms delay and 10mbits/s of bandwidth.
>
>
>
> I am running the default producer and consumer-with-timer
> from the ndn-cxx examples to randomly generate traffic
> flow between nodes. I am having an issue where, after a
> few seconds of only successful transfers, every consumer
> starts timing out on every request.
>
> When each consumer requests a packet every 100
> milliseconds, I get about 20 timeouts for every successful
> transfer. As I raise this period, things get better and by
> a period of 1 second, I get no timeouts. So it seems that
> there is some limitation to how much can be consumed or
> produced.
>
> So I am wondering, why does this happen? Is it a
> limitation of the consumer or the producer from ndn-cxx?
> Or is it related to the network emulation?
>
>
> Thanks in advance,
>
> André Dexheimer Carneiro
>
>
>
> _______________________________________________
> Mini-NDN mailing list
> Mini-NDN at lists.cs.ucla.edu <mailto:Mini-NDN at lists.cs.ucla.edu>
> http://www.lists.cs.ucla.edu/mailman/listinfo/mini-ndn
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/mini-ndn/attachments/20210421/eb67dd2d/attachment.html>
More information about the Mini-NDN
mailing list