Re: GSW and performance

Karthik Balaguru <[email protected]> Tue, 28 Apr 2009 18:57:45 +0530
Newsgroups gmane.ietf.rohc
Message-ID <[email protected]>
Hi,

In the case of LOSS option in Feedback packet,  if there is a large number
of
sequential packet drops, then the gsw size should be increased accordingly
to minimize
the losses at decompressor.

But,
If we increase the size to a large value (example - 50 or more), then it
will take
lot of time(processing cycles/cpu cycles) to operate(Encoding) on the gsw as
it
has to consider all the values inside the window .

So, what could be the optimal gsw size such that it does not affect the
performance(processing cycles/cpu cycles) and also the packet drops.

Thx in advans,
Karthik Balaguru

On Mon, Apr 27, 2009 at 8:31 PM, Carsten Bormann <[email protected]> wrote:

> Increasing the GSW size could reduce the performance.
>> So, What could be the optimal GSW size for good performance while
>> considering the Feedback options w.r.t the LOSS Option ?
>>
>
> Karthik,
>
> I don't think you will get good input from your two-liner questions.
> You need to explain more of the context you are coming from and what
> problem you are trying to solve.
> It may help to read:
>  http://www.catb.org/~esr/faqs/smart-questions.html<http://www.catb.org/%7Eesr/faqs/smart-questions.html>
>
> How much sliding window your compressor needs depends a lot on what your
> compressor does.
> Section 6 of RFC 3095 is about implementation issues, please read again:
>
> 6.  Implementation issues
>
>   This document specifies mechanisms for the protocol and leaves many
>   details on the use of these mechanisms to the implementers.  This
>   chapter is aimed to give guidelines, ideas and suggestions for
>   implementing the scheme.
>
> Section 5.7.6.9 tells you a lot about how the information in the LOSS
> option could be used by the compressor.
>
> Gruesse, Carsten
>
> _______________________________________________
> Rohc mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/rohc
>

_______________________________________________
Rohc mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/rohc