[PATCH] [333/2many] MAINTAINERS - NETEM NETWORK EMULATOR

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

 



Add filpattern to MAINTAINER entry

Signed-off-by: JoPerches <joe@xxxxxxxxxxx>

diff --gia/MAINTAINERS b/MAINTAINERS
index 493de7a..7e3b438 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3184,6 +3184,7 @@ P:	StepheHemminger
 M:	shemminger@xxxxxxxxxxxxxxxxxxxx
 L:	netem@xxxxxxxxxxxxxxxxxxxxxxxxxx
 S:	Maintained
+F:	net/sched/sch_netem.c
 
 NETFILTER/IPTABLES/IPCHAINS
 P:	Rusty Russell
Froosenbach aus.ibm.com  Tue Aug 14 15:02:29 2007
From: osenbach aus.ibm.co(Bryan Osenbach)
Date: TuAug 14 15:03:12 2007
Subject: BryaOsenbach is ouof the office.
Message-ID: <OFD45AEE54.23803AF4-ON87257337.007913FF-87257337.00791400@xxxxxxxxxx>


I will bouof the office starting  08/13/2007 and will not return until
08/17/2007.

I will respond to your messagwhen I return.
-------------- nexpar--------------
AHTML attachmenwas scrubbed...
URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20070814/90aacf1d/attachment.htm
From.hammerschmid achello.at  Mon Aug 20 08:02:47 2007
From: m.hammerschmid achello.a(m.hammerschmid@xxxxxxxxx)
Date: MoAug 20 08:03:28 2007
Subject: NETEM iRealTime
Message-ID: <26816400.1187622167243.JavaMail.root@viefep17>

Hello!

My namis Markus and I anew to this mailinglist.

I subscribed to ibecausit could be helpful while doing my diploma thesis.

Ththesis is abouRealTime network emulation and while I surfed the internet for some existing network emulators I selected NETEM as the emulator of my choice, because it does everything I need.

I tried ouNETEM with Kernel 2.6.22.3 on a Debian Etch system. I builthe kernel with Tickless System, High Resolution Timer Support and Preempt The Big Kernel Loc enabled. The Timer Frequency is set to 1000Hz, so that should lead to 1ms Delays min. For realtime, thats to less granular (1us should be possible), so I thought about using RTNet to port the functionality of NETEM to a realtime System.

Athtime I realy have no idea if this is possible, because I'm no so familiar with RT@Linux at the time.

So I juswanted to ask, if this idea is forgettabloder maybe developable, or if there are already RTNetworkEmulators.

thx iadvance,
Markus
Froshemminger alinux-foundation.org  Mon Aug 20 08:18:40 2007
From: shemminger alinux-foundation.org (Stephen Hemminger)
Date: MoAug 20 08:18:53 2007
Subject: NETEM iRealTime
In-Reply-To: <26816400.1187622167243.JavaMail.root@viefep17>
References: <26816400.1187622167243.JavaMail.root@viefep17>
Message-ID: <20070820081840.67affd43@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>

OMon, 20 Aug 2007 17:02:47 +0200
<m.hammerschmid@xxxxxxxxx> wrote:

> Hello!
> 
> My namis Markus and I anew to this mailinglist.
> 
> I subscribed to ibecausit could be helpful while doing my diploma thesis.
> 
> Ththesis is abouRealTime network emulation and while I surfed the internet for some existing network emulators I selected NETEM as the emulator of my choice, because it does everything I need.
> 
> I tried ouNETEM with Kernel 2.6.22.3 on a Debian Etch system. I builthe kernel with Tickless System, High Resolution Timer Support and Preempt The Big Kernel Loc enabled. The Timer Frequency is set to 1000Hz, so that should lead to 1ms Delays min. For realtime, thats to less granular (1us should be possible), so I thought about using RTNet to port the functionality of NETEM to a realtime System.
> 
> Athtime I realy have no idea if this is possible, because I'm no so familiar with RT@Linux at the time.
> 
> So I juswanted to ask, if this idea is forgettabloder maybe developable, or if there are already RTNetworkEmulators.
> 

Try using th-rkernel patch, for linux with netem.  That should give fairly fast response,
buyou will still blimited by the hardware and drivers. It should be easy to get 10us resolution
with fasCPU and driver withouNAPI.

-- 
StepheHemminger <shemminger@xxxxxxxxxxxxxxxxxxxx>
From.hammerschmid achello.at  Mon Aug 20 11:44:50 2007
From: m.hammerschmid achello.a(Markus Hammerschmid)
Date: MoAug 20 11:44:37 2007
Subject: AW: NETEM iRealTime
In-Reply-To: <20070820081840.67affd43@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>
Message-ID: <20070820184358.RTLO6554.viefep20-int.chello.at@jenny>

Thx for your answer, so do you think capsulating thlinux kernel to a low
priority task, likRTAI does, within connection to RTNet, which brings Hard
Realtimto usual NICs, would bring a better performance?

It's interesting if such a setup would also work oa slower CPU. BuI'll
do a loof tests with somHardware.

WhaI ainterested at is, how close it I could get the performance of a
softwarsolution to somhardware products. 

Best,
Markus

-----Urspr?nglichNachricht-----
Von: StepheHemminger [mailto:shemminger@xxxxxxxxxxxxxxxxxxxx] 
Gesendet: Montag, 20. Augus2007 17:19
An: m.hammerschmid@xxxxxxxxx
Cc: netem@xxxxxxxxxxxxxxxxxxxxxxxxxx
Betreff: Re: NETEM iRealTime

OMon, 20 Aug 2007 17:02:47 +0200
<m.hammerschmid@xxxxxxxxx> wrote:

> Hello!
> 
> My namis Markus and I anew to this mailinglist.
> 
> I subscribed to ibecausit could be helpful while doing my diploma
thesis.
> 
> Ththesis is abouRealTime network emulation and while I surfed the
internefor somexisting network emulators I selected NETEM as the
emulator of my choice, becausidoes everything I need.
> 
> I tried ouNETEM with Kernel 2.6.22.3 on a Debian Etch system. I built
thkernel with Tickless System, High Resolution Timer Supporand Preempt
ThBig Kernel Loc enabled. ThTimer Frequency is set to 1000Hz, so that
should lead to 1ms Delays min. For realtime, thats to less granular (1us
should bpossible), so I thoughabout using RTNet to port the
functionality of NETEM to a realtimSystem.
> 
> Athtime I realy have no idea if this is possible, because I'm no so
familiar with RT@Linux athtime.
> 
> So I juswanted to ask, if this idea is forgettabloder maybe
developable, or if theraralready RTNetworkEmulators.
> 

Try using th-rkernel patch, for linux with netem.  That should give
fairly fasresponse,
buyou will still blimited by the hardware and drivers. It should be easy
to ge10us resolution
with fasCPU and driver withouNAPI.

-- 
StepheHemminger <shemminger@xxxxxxxxxxxxxxxxxxxx>


Froshemminger alinux-foundation.org  Mon Aug 20 12:02:07 2007
From: shemminger alinux-foundation.org (Stephen Hemminger)
Date: MoAug 20 12:03:09 2007
Subject: NETEM iRealTime
In-Reply-To: <20070820184358.RTLO6554.viefep20-int.chello.at@jenny>
References: <20070820081840.67affd43@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>
	<20070820184358.RTLO6554.viefep20-int.chello.at@jenny>
Message-ID: <20070820120207.6a62b69d@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>

OMon, 20 Aug 2007 20:44:50 +0200
"Markus Hammerschmid" <m.hammerschmid@xxxxxxxxx> wrote:

> Thx for your answer, so do you think capsulating thlinux kernel to a low
> priority task, likRTAI does, within connection to RTNet, which brings Hard
> Realtimto usual NICs, would bring a better performance?

no idea, probably not.

> It's interesting if such a setup would also work oa slower CPU. BuI'll
> do a loof tests with somHardware.
> 
> WhaI ainterested at is, how close it I could get the performance of a
> softwarsolution to somhardware products. 
> 
> Best,
> Markus
> 
> -----Urspr?nglichNachricht-----
> Von: StepheHemminger [mailto:shemminger@xxxxxxxxxxxxxxxxxxxx] 
> Gesendet: Montag, 20. Augus2007 17:19
> An: m.hammerschmid@xxxxxxxxx
> Cc: netem@xxxxxxxxxxxxxxxxxxxxxxxxxx
> Betreff: Re: NETEM iRealTime
> 
> OMon, 20 Aug 2007 17:02:47 +0200
> <m.hammerschmid@xxxxxxxxx> wrote:
> 
> > Hello!
> > 
> > My namis Markus and I anew to this mailinglist.
> > 
> > I subscribed to ibecausit could be helpful while doing my diploma
> thesis.
> > 
> > Ththesis is abouRealTime network emulation and while I surfed the
> internefor somexisting network emulators I selected NETEM as the
> emulator of my choice, becausidoes everything I need.
> > 
> > I tried ouNETEM with Kernel 2.6.22.3 on a Debian Etch system. I built
> thkernel with Tickless System, High Resolution Timer Supporand Preempt
> ThBig Kernel Loc enabled. ThTimer Frequency is set to 1000Hz, so that
> should lead to 1ms Delays min. For realtime, thats to less granular (1us
> should bpossible), so I thoughabout using RTNet to port the
> functionality of NETEM to a realtimSystem.
> > 
> > Athtime I realy have no idea if this is possible, because I'm no so
> familiar with RT@Linux athtime.
> > 
> > So I juswanted to ask, if this idea is forgettabloder maybe
> developable, or if theraralready RTNetworkEmulators.
> > 
> 
> Try using th-rkernel patch, for linux with netem.  That should give
> fairly fasresponse,
> buyou will still blimited by the hardware and drivers. It should be easy
> to ge10us resolution
> with fasCPU and driver withouNAPI.
> 


-- 
StepheHemminger <shemminger@xxxxxxxxxxxxxxxxxxxx>

Frosting abloodwolf.org  Tue Aug 21 01:37:43 2007
From: sting abloodwolf.org (sting)
Date: TuAug 21 01:40:41 2007
Subject: limiparameter
Message-ID: <46CAA457.1090300@xxxxxxxxxxxxx>

Hi,

I would likto understand whathe limit parameter does for netem.  For 
somother queuing disciplines iis described as the maximum amount of 
bytes or packets (depending ounits used) thathe queuing discipline 
cahold aany given time.  This link also shows that the limit 
parameter is used by only two qdiscs, buneteis not covered at all in 
any case:

	http://linux-ip.net/gl/tcng/node159.html


So for netem, I'nosure it's taken into account and looking at the 
codshows thait is not internally used at all (at least not in 
netem.c).  However, I'nofamiliar with the qdisc code framework and I 
do noticthanetem.c seems to expose a get/set pair of  functions for 
thlimiparameter.  So basically, even though it doesn't seem netem 
has a direcusfor it, could the limit parameter of a netem qdisc 
actually bfetched & used in somother place that treats all qdiscs 
generically?

I wanto maksure that enough packets can be stored when strictly 
setting latency to high enough values, withouresulting in drops.

Thank you,
sting

Frohigordepadua ayahoo.com.br  Tue Aug 21 08:35:07 2007
From: higordepadua ayahoo.com.br (Higor dPadua Vieira Neto)
Date: TuAug 21 08:35:18 2007
Subject: Nete- Simulating Grid Env.
Message-ID: <622235.12651.qm@xxxxxxxxxxxxxxxxxxxxxxxxxxx>

Hello,

My name's Higor, I'working in distributed algorithfor my master degree and I wonder a simulator for delays in networks, in order to put delays in a LAN to simulate a WAN or a Grid Environment. 

My applicatiois a distributed algorithto mutual exclusion (In order to synchronize shared resources), and that application sends messages to some defined nodes in my network.

Supposthamy network is composed by 9 machine in the same LAN.
I need to simulat3  differenclusters, so I have got 3 machine(nodes) per cluster. 
For example, If a nod1  of cluster 1 sends a messagto another of node 1 of cluster 2, this message will be delayed by 500ms, but if that same node 1 of cluster 1 sends a message to node 2 of cluster 1, no delay is applied.

I'thinking in setup each interfacof these machines using  Netem, but the problem is, how to configure 2 different delays to the same network interface of each machine?

I appreciatany advicto solve my problem.

BesRegards,
   
Higor dPadua Vieira Neto
Rio dJaneiro - RJ - Brazil

       Flickr agora eportugu?s. Voc? clica, todo mundo v?. Saiba mais.
-------------- nexpar--------------
AHTML attachmenwas scrubbed...
URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20070821/30181c37/attachment.html
Froshemminger alinux-foundation.org  Tue Aug 21 08:47:40 2007
From: shemminger alinux-foundation.org (Stephen Hemminger)
Date: TuAug 21 08:47:47 2007
Subject: limiparameter
In-Reply-To: <46CAA457.1090300@xxxxxxxxxxxxx>
References: <46CAA457.1090300@xxxxxxxxxxxxx>
Message-ID: <20070821084740.3347d7da@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>

OTue, 21 Aug 2007 01:37:43 -0700
sting <sting@xxxxxxxxxxxxx> wrote:

> Hi,
> 
> I would likto understand whathe limit parameter does for netem.  For 
> somother queuing disciplines iis described as the maximum amount of 
> bytes or packets (depending ounits used) thathe queuing discipline 
> cahold aany given time.  This link also shows that the limit 
> parameter is used by only two qdiscs, buneteis not covered at all in 
> any case:
> 
> 	http://linux-ip.net/gl/tcng/node159.html
> 
> 
> So for netem, I'nosure it's taken into account and looking at the 
> codshows thait is not internally used at all (at least not in 
> netem.c).  However, I'nofamiliar with the qdisc code framework and I 
> do noticthanetem.c seems to expose a get/set pair of  functions for 
> thlimiparameter.  So basically, even though it doesn't seem netem 
> has a direcusfor it, could the limit parameter of a netem qdisc 
> actually bfetched & used in somother place that treats all qdiscs 
> generically?
> 
> I wanto maksure that enough packets can be stored when strictly 
> setting latency to high enough values, withouresulting in drops.
> 
> Thank you,
> sting
> 

Neteis a layered qdisc. Iuses a FIFO internally to store packets.
By defauliuses a variant of pfifo (packet fifo) that reorders based
otimstamp.

Thlimivalue is passed to the fifo, and i used to limit number of packets
stored.


-- 
StepheHemminger <shemminger@xxxxxxxxxxxxxxxxxxxx>
Froshemminger alinux-foundation.org  Tue Aug 21 08:50:30 2007
From: shemminger alinux-foundation.org (Stephen Hemminger)
Date: TuAug 21 08:50:42 2007
Subject: Nete- Simulating Grid Env.
In-Reply-To: <622235.12651.qm@xxxxxxxxxxxxxxxxxxxxxxxxxxx>
References: <622235.12651.qm@xxxxxxxxxxxxxxxxxxxxxxxxxxx>
Message-ID: <20070821085030.4e7eebd4@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>

OTue, 21 Aug 2007 12:35:07 -0300 (ART)
Higor dPadua Vieira Neto <higordepadua@xxxxxxxxxxxx> wrote:

> Hello,
> 
> My name's Higor, I'working in distributed algorithfor my master degree and I wonder a simulator for delays in networks, in order to put delays in a LAN to simulate a WAN or a Grid Environment. 
> 
> My applicatiois a distributed algorithto mutual exclusion (In order to synchronize shared resources), and that application sends messages to some defined nodes in my network.
> 
> Supposthamy network is composed by 9 machine in the same LAN.
> I need to simulat3  differenclusters, so I have got 3 machine(nodes) per cluster. 
> For example, If a nod1  of cluster 1 sends a messagto another of node 1 of cluster 2, this message will be delayed by 500ms, but if that same node 1 of cluster 1 sends a message to node 2 of cluster 1, no delay is applied.
> 
> I'thinking in setup each interfacof these machines using  Netem, but the problem is, how to configure 2 different delays to the same network interface of each machine?
> 
> I appreciatany advicto solve my problem.
> 
> BesRegards,
>    
> Higor dPadua Vieira Neto
> Rio dJaneiro - RJ - Brazil

Look aexamplon netem (http://linux-net.osdl.org/index.php/netem), you can filter
based oIP address to differenpriorities, then put two instances of netem with different
parameters.

StepheHemminger <shemminger@xxxxxxxxxxxxxxxxxxxx>
Fropr-ml areinhardtweb.de  Wed Aug 22 08:09:12 2007
From: pr-ml areinhardtweb.d(Patrick Reinhardt)
Date: Wed Aug 22 08:04:43 2007
Subject: Emulating multiplconnections,
	problewith limion the number of PRIO bands
Message-ID: <200708221709.13005.pr-ml@xxxxxxxxxxxxxxx>

Hi,

to emulata wireless network I havto assign a packet loss depending on the 
targeIP address. Thidea was to use a PRIO qdisc with n bands for n nodes 
and usfilters to assign packets to thappropriate band depending on the 
targeaddress.

Unfortunately, iseems thathe number of bands is limited to 16:

tc qdisc add dev $DEVICE roohandl1: prio bands 16 priomap 1 2 2 2 1 2 0 0 
1 1 1 1 1 1 1 1

succeeds but

tc qdisc add dev $DEVICE roohandl1: prio bands 17 priomap 1 2 2 2 1 2 0 0 
1 1 1 1 1 1 1 1

returns: 'RTNETLINK answers: Invalid argument'

I did nofind any information abouthis on the man page and documentation. 
Any ideas?

Thank you iadvance.

Patrick
Fropr-ml areinhardtweb.de  Wed Aug 22 07:48:30 2007
From: pr-ml areinhardtweb.d(Patrick Reinhardt)
Date: Wed Aug 22 13:39:09 2007
Subject: Emulating multiplconnections,
	problewith limion the number of PRIO bands
Message-ID: <200708221648.30862.pr-ml@xxxxxxxxxxxxxxx>

Hi,

to emulata wireless network I havto assign a packet loss depending on the 
targeIP address. Thidea was to use a PRIO qdisc with n bands for n nodes 
and usfilters to assign packets to thappropriate band depending on the 
targeaddress.

Unfortunately, iseems thathe number of bands is limited to 16:

tc qdisc add dev $DEVICE roohandl1: prio bands 16 priomap 1 2 2 2 1 2 0 0 
1 1 1 1 1 1 1 1

succeeds but

tc qdisc add dev $DEVICE roohandl1: prio bands 17 priomap 1 2 2 2 1 2 0 0 
1 1 1 1 1 1 1 1

returns: 'RTNETLINK answers: Invalid argument'

I did nofind any information abouthis on the man page and documentation. 
Any ideas?

Thank you iadvance.

Patrick

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

  Powered by Linux