GSoC on high performance ns-3 network emulation

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

GSoC on high performance ns-3 network emulation

Pasquale Imputato
Hi everyone,

I’m a PhD student at University of Naples; I already contributed to ns-3
through design and implementation of the queueing disciplines in
traffic-control module.

I plan to submit a proposal for GSoC on high performance ns-3 network
emulation reported in the ns-3 GSoC ideas page mentored by Stefano Avallone.

The proposal project intends to improve the current poor emulation
performances. The proposed approach aims to remove the intensive polling
between user and kernel space by the introduction of netmap in ns-3. Also,
it designs and introduces a device flow control on a memory mapped device
to make traffic-control aware of the device queue status and enable a more
accurate evaluation of the scheduling algorithms in emulated scenarios.
Finally, it aims to introduce the required optimization to take advantages
of the changes introduced. The projects extend evaluate the changes
introduced on real device with an accurate performance evaluation.

Follow the proposed plan:
Phase 1: netmap integration in ns-3, introduction of new device and its
helper that uses netmap primitives, testing;
Phase 2: design and implementation of the flow control mechanism on a
memory mapped device, testing, initial performance evaluation, initial
documentation, addressing reviews come from the Phase 1;
Phase 3: implementation of the required optimization to take advantages of
the changes introduced, testing, performance evaluation, documentation,
refinement of the whole project, addressing reviews come from the Phase 2.

Any comments and proposal reviews are more than welcome.

Thanks,

Pasquale Imputato
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: GSoC on high performance ns-3 network emulation

Tom Henderson-2
On 03/24/2017 09:45 AM, Pasquale Imputato wrote:

> Hi everyone,
>
> I’m a PhD student at University of Naples; I already contributed to ns-3
> through design and implementation of the queueing disciplines in
> traffic-control module.
>
> I plan to submit a proposal for GSoC on high performance ns-3 network
> emulation reported in the ns-3 GSoC ideas page mentored by Stefano Avallone.
>
> The proposal project intends to improve the current poor emulation
> performances. The proposed approach aims to remove the intensive polling
> between user and kernel space by the introduction of netmap in ns-3. Also,
> it designs and introduces a device flow control on a memory mapped device
> to make traffic-control aware of the device queue status and enable a more
> accurate evaluation of the scheduling algorithms in emulated scenarios.
> Finally, it aims to introduce the required optimization to take advantages
> of the changes introduced. The projects extend evaluate the changes
> introduced on real device with an accurate performance evaluation.
>
> Follow the proposed plan:
> Phase 1: netmap integration in ns-3, introduction of new device and its
> helper that uses netmap primitives, testing;
> Phase 2: design and implementation of the flow control mechanism on a
> memory mapped device, testing, initial performance evaluation, initial
> documentation, addressing reviews come from the Phase 1;
> Phase 3: implementation of the required optimization to take advantages of
> the changes introduced, testing, performance evaluation, documentation,
> refinement of the whole project, addressing reviews come from the Phase 2.
>
> Any comments and proposal reviews are more than welcome.
>
> Thanks,
>
> Pasquale Imputato

Hi Pasquale, it looks like it will be an interesting proposal. Please
try to cover these points:

- why netmap vs. another framework such as PF_RING?  What will be the
underlying system requirements to use it?

- what performance specifically are you trying to address; when you say
'poor', can you quantify?

- what performance do you expect to be achievable, can you quantify?

- how will it be benchmarked?  Is special hardware needed to benchmark?  
What kind of tracing support from within the model can be used to benchmark?

Thanks,

Tom


Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: GSoC on high performance ns-3 network emulation

Pasquale Imputato
Hi Tom,

thank you for your comments. I submitted the draft on the GSoC project page.
I tried to cover the points in the draft.

Any comments are much appreciated.

Thanks,

Pasquale Imputato

2017-03-25 15:43 GMT+01:00 Tom Henderson <[hidden email]>:

> On 03/24/2017 09:45 AM, Pasquale Imputato wrote:
>
>> Hi everyone,
>>
>> I’m a PhD student at University of Naples; I already contributed to ns-3
>> through design and implementation of the queueing disciplines in
>> traffic-control module.
>>
>> I plan to submit a proposal for GSoC on high performance ns-3 network
>> emulation reported in the ns-3 GSoC ideas page mentored by Stefano
>> Avallone.
>>
>> The proposal project intends to improve the current poor emulation
>> performances. The proposed approach aims to remove the intensive polling
>> between user and kernel space by the introduction of netmap in ns-3. Also,
>> it designs and introduces a device flow control on a memory mapped device
>> to make traffic-control aware of the device queue status and enable a more
>> accurate evaluation of the scheduling algorithms in emulated scenarios.
>> Finally, it aims to introduce the required optimization to take advantages
>> of the changes introduced. The projects extend evaluate the changes
>> introduced on real device with an accurate performance evaluation.
>>
>> Follow the proposed plan:
>> Phase 1: netmap integration in ns-3, introduction of new device and its
>> helper that uses netmap primitives, testing;
>> Phase 2: design and implementation of the flow control mechanism on a
>> memory mapped device, testing, initial performance evaluation, initial
>> documentation, addressing reviews come from the Phase 1;
>> Phase 3: implementation of the required optimization to take advantages of
>> the changes introduced, testing, performance evaluation, documentation,
>> refinement of the whole project, addressing reviews come from the Phase 2.
>>
>> Any comments and proposal reviews are more than welcome.
>>
>> Thanks,
>>
>> Pasquale Imputato
>>
>
> Hi Pasquale, it looks like it will be an interesting proposal. Please try
> to cover these points:
>
> - why netmap vs. another framework such as PF_RING?  What will be the
> underlying system requirements to use it?
>
> - what performance specifically are you trying to address; when you say
> 'poor', can you quantify?
>
> - what performance do you expect to be achievable, can you quantify?
>
> - how will it be benchmarked?  Is special hardware needed to benchmark?
> What kind of tracing support from within the model can be used to benchmark?
>
> Thanks,
>
> Tom
>
>
>
Loading...