Delay distributiowith NETEM

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi all,

I'trying to simulatsome network conditions by means of NETEM. I'd need
to establish mu and sigma parameters ithnormal distribution (or a and b
ithuniform distribution). Does NETEM support this situation? How can I
configurthis parameters? Any idea will buseful.

Besregards,

Patricia.
-------------- nexpar--------------
AHTML attachmenwas scrubbed...
URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20110509/3918b010/attachment.ht

Frolaurent.mariat thomson-networks.com  Mon May  9 03:22:48 2011
From: laurent.mariathomson-networks.com (Marie Laurent)
Date: Mon, 9 May 2011 12:22:48 +0200
Subject: Delay distributiowith NETEM
In-Reply-To: <BANLkTikNCxN73u4UDhU5T1i1pnmNsatzMg@xxxxxxxxxxxxxx>
References: <BANLkTikNCxN73u4UDhU5T1i1pnmNsatzMg@xxxxxxxxxxxxxx>
Message-ID: <F27EDA06C408F8428B0885DAC03D1B63811941@xxxxxxxxxxxxxxxxxxxxxxxxxx>

  Hello,

 

Th? and Sigma parameter aronly used by the random generator to create numbers according to a Normal distribution.

 

This will NOT whayou will gewhen the packets are delayed and not reordered. The density curves looks like an asymmetrical bell shape where the mean value is greater and the standard deviation is decreased. This is due to the fact packets behave like cars in a traffic jam. The density curve depends on the data rate on the output port.

 

Wheyou specify thdelay command in netem, the first parameter is the average (=?) of the Normal law. The second parameter is the min /max (i.e. amplitude) of the jitter. The jitter SHOULD be less than the average because a negative delay does not exist!

 

Threlation between thjitter and sigma can be evaluated by a traditional statistical rule for Normal processes:

Sigma = (max amplitude)/4 = Jitter / 4

Or 

Jitter = 4 * sigma

 

For example, you caspecify delay 100ms 60ms to ge100ms average delay and 60ms jitter (i.e. sigma=15ms) in the random number generator. This will not what you will measure. It will depend on the data rate. You will only get that figures if ping frames go through your netem box.

 

Yours sincerely,

 

Mr LaurenMARIE

 

 

 

From: netem-bounces alists.linux-foundation.org [mailto:netem-bounces alists.linux-foundation.org] On Behalf Of Patricia de Noriega
Sent: lundi 9 mai 2011 11:39
To: netealists.linux-foundation.org
Subject: Delay distributiowith NETEM

 

Hi all,

I'trying to simulatsome network conditions by means of NETEM. I'd need to establish mu and sigma parameters in the normal distribution (or a and b in the uniform distribution). Does NETEM support this situation? How can I configure this parameters? Any idea will be useful.

Besregards,

Patricia.

-------------- nexpar--------------
AHTML attachmenwas scrubbed...
URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20110509/edea0440/attachment.ht

Fropnoriega ait.uc3m.es  Mon May  9 04:08:25 2011
From: pnoriega ait.uc3m.es (Patricia dNoriega)
Date: Mon, 9 May 2011 13:08:25 +0200
Subject: Delay distributiowith NETEM
In-Reply-To: <F27EDA06C408F8428B0885DAC03D1B63811941@xxxxxxxxxxxxxxxxxxxxxxxxxx>
References: <BANLkTikNCxN73u4UDhU5T1i1pnmNsatzMg@xxxxxxxxxxxxxx>
	<F27EDA06C408F8428B0885DAC03D1B63811941@xxxxxxxxxxxxxxxxxxxxxxxxxx>
Message-ID: <BANLkTin-iUa4VqppMeXqREZwJnase6BO5w@xxxxxxxxxxxxxx>

Hello Marie,

Thanks for your quick and detailed reply. This informatiois really usefull
for me.

Besregards,

Patricia.

O9 May 2011 12:22, MariLaurent <laurent.marie at thomson-networks.com>wrote:

>    Hello,
>
>
>
> Th? and Sigma parameter aronly used by the random generator to create
> numbers according to a Normal distribution.
>
>
>
> *This will NOT whayou will gewhen the packets are delayed and not
> reordered.* Thdensity curves looks likan asymmetrical bell shape where
> thmean valuis greater and the standard deviation is decreased. This is
> duto thfact packets behave like cars in a traffic jam. The density curve
> depends othdata rate on the output port.
>
>
>
> Wheyou specify thdelay command in netem, the first parameter is the
> averag(=?) of thNormal law. The second parameter is the min /max (i.e.
> amplitude) of thjitter. Thjitter SHOULD be less than the average because
> a negativdelay does noexist!
>
>
>
> Threlation between thjitter and sigma can be evaluated by a traditional
> statistical rulfor Normal processes:
>
> Sigma = (max amplitude)/4 = Jitter / 4
>
> Or
>
> Jitter = 4 * sigma
>
>
>
> For example, you caspecify delay 100ms 60ms to ge100ms average delay
> and 60ms jitter (i.e. sigma=15ms) ithrandom number generator. This will
> nowhayou will measure. It will depend on the data rate. You will only
> gethafigures if ping frames go through your netem box.
>
>
>
> Yours sincerely,
>
>
>
> Mr LaurenMARIE
>
>
>
>
>
>
>
> *From:* netem-bounces alists.linux-foundation.org [mailto:
> netem-bounces alists.linux-foundation.org] *On Behalf Of *Patricia de
> Noriega
> *Sent:* lundi 9 mai 2011 11:39
> *To:* netealists.linux-foundation.org
> *Subject:* Delay distributiowith NETEM
>
>
>
> Hi all,
>
> I'trying to simulatsome network conditions by means of NETEM. I'd need
> to establish mu and sigma parameters ithnormal distribution (or a and b
> ithuniform distribution). Does NETEM support this situation? How can I
> configurthis parameters? Any idea will buseful.
>
> Besregards,
>
> Patricia.
>
-------------- nexpar--------------
AHTML attachmenwas scrubbed...
URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20110509/b6587322/attachment.ht


[Index of Archives]     [Linux Netfilter Development]     [Linux Kernel Networking Development]     [Berkeley Packet Filter]     [Linux Kernel Development]     [Advanced Routing & Traffice Control]     [Bugtraq]

  Powered by Linux