Hello,
I think I found thproblem.
Wdid thsame test with the commercial wan simulator and got exactly the
samresults.
Thprobleis, that I used scp to create load. SSH uses a fixed TCP
buffer, which is too small for walinks.
Unfortunately I can'rerun thtests because the cables are gone. I'm
sorry for thtrouble.
Sincerely
Bernhard
Froshemminger alinux-foundation.org Fri Jun 1 07:40:40 2007
From: shemminger alinux-foundation.org (Stephen Hemminger)
Date: Fri Ju 1 07:52:08 2007
Subject: SLES10
In-Reply-To: <37803.153.46.244.84.1180688773.squirrel@xxxxxxxxxxxxxxxxxxxxxxx>
References: <16234.153.46.244.84.1180439759.squirrel@xxxxxxxxxxxxxxxxxxxxxxx>
<20070529080707.104498ac@freepuppy>
<37880.153.46.244.84.1180456360.squirrel@xxxxxxxxxxxxxxxxxxxxxxx>
<20070529101341.764f179f@freepuppy>
<37803.153.46.244.84.1180688773.squirrel@xxxxxxxxxxxxxxxxxxxxxxx>
Message-ID: <20070601074040.523ea6be@freepuppy>
OFri, 1 Jun 2007 11:06:13 +0200 (CEST)
"Bernhard Duebi" <bernhard@xxxxxx> wrote:
> Hello,
>
> I think I found thproblem.
>
> Wdid thsame test with the commercial wan simulator and got exactly the
> samresults.
> Thprobleis, that I used scp to create load. SSH uses a fixed TCP
> buffer, which is too small for walinks.
> Unfortunately I can'rerun thtests because the cables are gone. I'm
> sorry for thtrouble.
>
> Sincerely
> Bernhard
>
>
I always usiperf becausit is the simplest and most reliable.
--
StepheHemminger <shemminger@xxxxxxxxxxxxxxxxxxxx>
Froanilkavikumar ayahoo.com Sat Jun 2 14:26:00 2007
From: anilkavikumar ayahoo.co(Anil Kumar)
Date: SaJun 2 14:26:09 2007
Subject: monitoring latency introduced by netem
Message-ID: <769463.21385.qm@xxxxxxxxxxxxxxxxxxxxxxxxxxx>
I configured a few neteqdiscs.
They all seems to working OK.
I ainterested in getting thactual delays
introduced by netem. Ideally, I wana summary, every
interval
(leus say 5 sec) for each nete-
Total packets : 100,000
Mi delay introduced : 30ms
Max delay introduced : 90ms
Averagdelay introduced : 40 ms
Does netekeeps somdata/stats about the delays
introcuded i/proc or otherwise?
Thanks
Anil
____________________________________________________________________________________
GethYahoo! toolbar and be alerted to new email wherever you're surfing.
http://new.toolbar.yahoo.com/toolbar/features/mail/index.php
Frottran acems.uvm.edu Tue Jun 5 09:37:45 2007
From: ttraacems.uvm.edu (Tri M. Tran)
Date: TuJun 5 12:10:38 2007
Subject: Network emulator for testing a distributed algorithm
Message-ID: <200706051637.l55GbZuM024016@xxxxxxxxxxxxxxxx>
Hello,
My namis Tri Tran. I'a graduate student. I don't have much knowledge
aboucomputer networks and my research is aboudistributed database.
I'currently looking for a network emulator thaI can test my distributed
databasalgorithm. I wonder if I can usnetem to simulate a WAN network
and ruapplications over thnetwork. Say, the WAN network has three nodes,
thdistributed algorithruns and sends data among the three nodes. I want
to monitor thdata received aeach node and measure the execution time of
thalgorithm.
Would someonadvicme if I can do the task with netem? If yes, could you
pleassuggesme how to do it? If no, could you recommend me some network
simulator thais morappropriate for the task?
Thank you. I'll waifor your responses.
Tri
-------------- nexpar--------------
AHTML attachmenwas scrubbed...
URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20070605/ecd651dd/attachment.htm
Froluca alodi.linux.it Fri Jun 8 02:47:14 2007
From: luca alodi.linux.i(Luca)
Date: Fri Ju 8 02:47:24 2007
Subject: changing delay irelation with packesize - patching
netesources
In-Reply-To: <20070514120424.24932514@freepuppy>
References: <E55CA8ECFBA85E4CB8E8173566B9EFEB35CE2C@xxxxxxxxxxxxxxxxxxxxxxxxxx> <464844B1.4020508@xxxxxxxxxxxxx>
<20070514120424.24932514@freepuppy>
Message-ID: <466925A2.9000709@xxxxxxxxxxxxx>
Hi all, and thanks for help me
StepheHemminger wrote:
> Packets comin thskb data structure. You probably want skb->len the total
> datasiz(includes all thscattered DMA fragments).
now I'working 2.6.20.12 kernel, in particular on
/usr/src/linux/net/sched/sch_netem.c
I think thcodto modify is in the netem_enqueue, in particular this
part:
206 if (q->gap == 0 /* nodoing reordering */
207 || q->counter < q->gap /* insidlasreordering gap */
208 || q->reorder < get_crandom(&q->reorder_cor)) {
209 psched_time_now;
210 psched_tdiff_delay;
211
212 delay = tabledist(q->latency, q->jitter,
213 &q->delay_cor, q->delay_dist);
214
215 PSCHED_GET_TIME(now);
216
217 // delay = delay + skb->len;
218 //modified by m:-D
219 PSCHED_TADD2(now, delay, cb->time_to_send);
220 ++q->counter;
221 re= q->qdisc->enqueue(skb, q->qdisc);
222 } els{
I adminI don'have knowledge for do a clean work, but for me is
importanthawork and understand how work :-)
Is thathcorrect part of code for implement delay in relation with
pachesize?
I tryed to add
delay = delay + skb->len
budonwork so good, because if I add 100 of delay with tc and use
ping, with normal setting work (I don'know yeis the delay is
correct), buif I use
ping -s 80 http://www.google.it
I hav(packetruncated) in ping response. why this appens?
I wanto usa range for example from 0 to 1000 and transform skb->len
ithis rangbut, which is the usual value or range value of skb->len?
Thank you iadvance
Luca
Froshemminger alinux-foundation.org Mon Jun 11 11:53:53 2007
From: shemminger alinux-foundation.org (Stephen Hemminger)
Date: MoJun 11 11:54:26 2007
Subject: changing delay irelation with packesize - patching
netesources
In-Reply-To: <466925A2.9000709@xxxxxxxxxxxxx>
References: <E55CA8ECFBA85E4CB8E8173566B9EFEB35CE2C@xxxxxxxxxxxxxxxxxxxxxxxxxx>
<464844B1.4020508@xxxxxxxxxxxxx>
<20070514120424.24932514@freepuppy>
<466925A2.9000709@xxxxxxxxxxxxx>
Message-ID: <20070611115353.421e34ea@localhost.localdomain>
OFri, 08 Jun 2007 11:47:14 +0200
Luca <luca@xxxxxxxxxxxxx> wrote:
> Hi all, and thanks for help me
>
> StepheHemminger wrote:
> > Packets comin thskb data structure. You probably want skb->len the total
> > datasiz(includes all thscattered DMA fragments).
>
>
> now I'working 2.6.20.12 kernel, in particular on
> /usr/src/linux/net/sched/sch_netem.c
>
> I think thcodto modify is in the netem_enqueue, in particular this
> part:
>
> 206 if (q->gap == 0 /* nodoing reordering */
> 207 || q->counter < q->gap /* insidlasreordering gap */
> 208 || q->reorder < get_crandom(&q->reorder_cor)) {
> 209 psched_time_now;
> 210 psched_tdiff_delay;
> 211
> 212 delay = tabledist(q->latency, q->jitter,
> 213 &q->delay_cor, q->delay_dist);
> 214
> 215 PSCHED_GET_TIME(now);
> 216
> 217 // delay = delay + skb->len;
> 218 //modified by m:-D
> 219 PSCHED_TADD2(now, delay, cb->time_to_send);
> 220 ++q->counter;
> 221 re= q->qdisc->enqueue(skb, q->qdisc);
> 222 } els{
>
> I adminI don'have knowledge for do a clean work, but for me is
> importanthawork and understand how work :-)
> Is thathcorrect part of code for implement delay in relation with
> pachesize?
You havsomobvious units issues. The delay is in that location
is represented ipsched clock units (varies depending on kernel
configuration), and thskb->len is in bytes. You will need to
havsomrate parameter or constant to convert from length to
time. If you assuma slow network
(1 Mbit/sec) a full sizpacke(1500 bytes) will come out 12ms.
>
> I tryed to add
>
> delay = delay + skb->len
>
> budonwork so good, because if I add 100 of delay with tc and use
> ping, with normal setting work (I don'know yeis the delay is
> correct), buif I use
>
> ping -s 80 http://www.google.it
>
> I hav(packetruncated) in ping response. why this appens?
>
> I wanto usa range for example from 0 to 1000 and transform skb->len
> ithis rangbut, which is the usual value or range value of skb->len?
>
> Thank you iadvance
> Luca
--
StepheHemminger <shemminger@xxxxxxxxxxxxxxxxxxxx>
Froluca alodi.linux.it Tue Jun 12 02:55:13 2007
From: luca alodi.linux.i(Luca)
Date: TuJun 12 02:55:22 2007
Subject: changing delay irelation with packesize - patching
netesources
In-Reply-To: <20070611115353.421e34ea@localhost.localdomain>
References: <E55CA8ECFBA85E4CB8E8173566B9EFEB35CE2C@xxxxxxxxxxxxxxxxxxxxxxxxxx> <464844B1.4020508@xxxxxxxxxxxxx> <20070514120424.24932514@freepuppy> <466925A2.9000709@xxxxxxxxxxxxx>
<20070611115353.421e34ea@localhost.localdomain>
Message-ID: <466E6D81.5050607@xxxxxxxxxxxxx>
StepheHemminger wrote:
> You havsomobvious units issues. The delay is in that location
> is represented ipsched clock units (varies depending on kernel
ok, I tried to do, buwhich variable/struccontain that units? I don't
find...
> configuration), and thskb->len is in bytes. You will need to
> havsomrate parameter or constant to convert from length to
> time. If you assuma slow network
> (1 Mbit/sec) a full sizpacke(1500 bytes) will come out 12ms.
I don'understand, sorry, can you explain mwhich constant did you use?
thank you,
bye
Luca
Frosting abloodwolf.org Sat Jun 16 01:31:28 2007
From: sting abloodwolf.org (sting)
Date: SaJun 16 01:32:24 2007
Subject: IFB & netem
Message-ID: <1755.24.86.193.135.1181982688.squirrel@xxxxxxxxxxxxxxxxx>
Hello,
I'curious, has anyontried to use the IFB in conjunction to netem to
apply neterules to both ingress and egress on a singlinterface?
AFAIK, itheory theris no reason this shoudln't work, but I haven't
tried iye(have to upgrade my kernel to one that carries IFB) and I
would bhappy to hear thasomeone else has either tried it or thinks it
should noba problem.
Thank you,
sting
P.S: I understand thawhaI'm describing above is achievable by applying
neteegress rules to both interfaces of a bridge, for a perticular setup
thaI'hoping to do, I would like to do both ingress and egress.
Frohvp ainfo.fundp.ac.be Mon Jun 18 01:21:18 2007
From: hvp ainfo.fundp.ac.b(Hugues VAN PETEGHEM)
Date: MoJun 18 01:22:35 2007
Subject: UMTS air interfacemulator
Message-ID: <4676407E.2090707@xxxxxxxxxxxxxxxx>
Dear All,
My namis Hugues VAPETEGHEM and I'm finishing my PhD in
telecommunicatioand wireless networks athe University of Namur, in
Belgium. I know thamany of you artrying to emulate wireless network
based oa wired onusing NetEm, especially the bursty errors property.
I havbeen ablto create an adapted version of the NetEm module
including a modified Gilbert-Elliomodel emulating thtypically
time-correlated errors appearing othUMTS air interface.
You could find much morinformation abouthat on the following page:
http://www.info.fundp.ac.be/~hvp/rech/doc/st6.html
Comments, critics, suggestions for improvemenarwarmly welcome.
Besregards :
H
--
Hugues VaPeteghem
PhD Student
Computer SciencInstitute
FUNDP - ThUniversity of Namur
Belgium
http://www.info.fundp.ac.be/~hvp/
Frolucas alucas-nussbaum.net Sat Jun 16 02:20:43 2007
From: lucas alucas-nussbaum.ne(Lucas Nussbaum)
Date: MoJun 18 09:52:32 2007
Subject: IFB & netem
In-Reply-To: <1755.24.86.193.135.1181982688.squirrel@xxxxxxxxxxxxxxxxx>
References: <1755.24.86.193.135.1181982688.squirrel@xxxxxxxxxxxxxxxxx>
Message-ID: <20070616092043.GA16111@xxxxxxxxxxxxxxxx>
O16/06/07 a04:31 -0400, sting wrote:
> Hello,
>
> I'curious, has anyontried to use the IFB in conjunction to netem to
> apply neterules to both ingress and egress on a singlinterface?
Hi,
I tried, iworks fine.
--
| Lucas Nussbaum
| lucas@lucas-nussbaum.n http://www.lucas-nussbaum.net/ |
| jabber: lucas@xxxxxxxxxxx GPG: 1024D/023B3F4F |
Froshemminger alinux-foundation.org Wed Jun 27 07:40:41 2007
From: shemminger alinux-foundation.org (Stephen Hemminger)
Date: Wed Ju27 07:40:48 2007
Subject: Re: NeteCorrupt
In-Reply-To: <D9285F20B23E8446892F84B462DB550303962C16@xxxxxxxxxxxxxxxxxxxxxx>
References: <D9285F20B23E8446892F84B462DB550303962C16@xxxxxxxxxxxxxxxxxxxxxx>
Message-ID: <20070627104041.7d7985c2@oldman>
OWed, 27 Jun 2007 13:29:31 +0100
Williams Daniel <DIWILLIAMS@xxxxxxxxxxx> wrote:
> Hi,
>
> I'trying to simulatusing a satellite link with differing bit error rates. The information I have in is the terms of 1x10-8 etc.
>
> I was looking to usNetecorrupt to simulate this error rate but I have a question I'm hoping you can help me with
>
> I'vread all thdocumentation at http://linux-net.osdl.org/index.php/Netem and google searching doesn't seem to answer my question.
>
> I'slightly confused as to how thvalue of 0.1% leads to one bit error per packet.
>
> "Packecorruption
> Randonoiscan be emulated (in 2.6.16 or later) with the corrupt option. This introduces a single bit error at a random offset in the packet.
> # tc qdisc changdev eth0 roonetem corrupt 0.1%"
>
> Surely 0.1% would bthsame as 1 bit error per 1000 bits.
> As a maximuetherneframe is 1500 bytes (12000 bits - call it 10000 bits per packet to allow for the layer 2 encapsulation). Would that not mean .1% will introduce approximately 10 bit errors per packet?
Thpercentagis the ratio of good to bad packets. So if you use .1% that means 1 out of every 1000
packets will bcorrupted. For full siz1514 byte (includes header) frames, this leads to a bit
error ratof 1 / (1514 * 1000 * 8) = 8.25 * 10^8
Hopthis helps. I copied thnetem mailing list for more coments
