[Codel] Codel Digest, Vol 29, Issue 4
Kathleen Nichols
nichols at pollere.com
Sun Feb 22 16:25:52 EST 2015
I used TCP/Linux because it was closer to what is out there in a lot of
boxes
that people use and I think there may have been some ns-2 reason also but
I'm thankfully not using simulators currently, spending more time with
traces
and lab stuff. You can and should use anything you want.
Reverse ftps are used to create a flow of acks in the the monitored
direction. This
can make a huge difference in what you observe. I believe the first work
on this might
be something Lixia Zhang published as a grad student and I also found
this in early
simulation work I did. You could read about it and you can also do
different experiments
yourself to see how results differ. You might also read
http://queue.acm.org/detail.cfm?id=2209336 and some of the references cited.
On 2/22/15 6:35 AM, kanu monga wrote:
> Thank you so much for replying. i really appreciate it.
> i have two more questions:
> (1) what is the concept of using TCP/Linux with codel. can't we use
> other TCP Variants
> (2) N what is purpose of using reverse ftps
>
>
> On Fri, Feb 20, 2015 at 11:32 PM, <codel-request at lists.bufferbloat.net
> <mailto:codel-request at lists.bufferbloat.net>> wrote:
>
> Send Codel mailing list submissions to
> codel at lists.bufferbloat.net <mailto:codel at lists.bufferbloat.net>
>
> To subscribe or unsubscribe via the World Wide Web, visit
> https://lists.bufferbloat.net/listinfo/codel
> or, via email, send a message with subject or body 'help' to
> codel-request at lists.bufferbloat.net
> <mailto:codel-request at lists.bufferbloat.net>
>
> You can reach the person managing the list at
> codel-owner at lists.bufferbloat.net
> <mailto:codel-owner at lists.bufferbloat.net>
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of Codel digest..."
>
>
> Today's Topics:
>
> 1. Re: Codel code (Kathleen Nichols)
> 2. Re: Codel code (Dave Taht)
> 3. Setpoint in CoDel (Anjali Chawla)
> 4. Re: Setpoint in CoDel (Dave Taht)
> 5. Re: Setpoint in CoDel (Dave Taht)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Thu, 19 Feb 2015 13:34:14 -0800
> From: Kathleen Nichols <nichols at pollere.com
> <mailto:nichols at pollere.com>>
> To: codel at lists.bufferbloat.net <mailto:codel at lists.bufferbloat.net>
> Subject: Re: [Codel] Codel code
> Message-ID: <54E656D6.5010308 at pollere.com
> <mailto:54E656D6.5010308 at pollere.com>>
> Content-Type: text/plain; charset=windows-1252
>
>
> I'll reply list in case anyone else is interested. Just delete if you
> are fortunate enough to
> have nothing to do with ns-2.
>
> First of all, I'm as much a victim of ns-2's tcl as the next person and
> secondly, I will
> tell you what I remember without looking at the c++ code, but would
> recommend
> you take a look at that for some of the settings.
>
> The first commented line just shows how you can set stuff on the command
> line and what
> the different variables are. This particular script is using PackMime to
> generate web
> traffic though there is also an update of a web script I wrote many
> years ago and
> CableLabs paid me to update. I also used that for some codel tests. So
> this can run
> a mix of f ftp connections, w web connections per sec, c constant bit
> rate connections
> through a bottleneck of b Mbps. Use s to set the file size that the ftps
> transfer and if
> set to -1 the ftps are infinite, that is don't terminate over the
> simulation run. This avoids
> lots of slow starts. d is something I used to get the dynamic bandwidth
> thing to work,
> that is it makes the bottleneck link change periodically (you can see
> the plot of this
> in the ACM CoDel paper). Reverse ftps are sourced at the opposite end of
> the bottleneck
> so they create acks in the "forward" direction.
>
> There is a routine called build_cbr and that's where the rate and packet
> size for the CBRs is set. This sets dynamic bw to 0 so the bandwidth
> doesn't change unless set in the command line to do so. I don't remember
> off hand what greedy is. I think it sets one
> ftp to be "greedy".
>
> A bunch of this you should try to figure out for yourself by looking at
> ns2 documentation
> and perhaps at the agent .cc files.
>
> The nominal RTT is what is "expected" and is used as the interval but
> the real rtt is what
> is actually experienced in the particular experiment. I will use this
> opportunity once again
> to note that CoDel is actually more sensitive to this parameter than to
> target in the sense
> that excess drops will happen if the real rtt is longer than the
> interval (I keep trying to
> explain this but no one hears it.)
>
> Code to set the random seed. Sometimes you want different runs to use
> different random
> seeds and sometimes you want to run exactly the same scenario.
>
> There's a bunch of parameters to the TCP agents. Some of these are in
> ns2 documentation,
> but you can find it in the code also.
>
> Just a note that I couldn't get the ns2 TCP/Linux code to work the way
> it was supposed to,
> specifically the idle() routine never seemed to get called. (This sort
> of thing is pretty
> common with ns2) and I didn't want to waste a lot of time untangling the
> tcl-c++
> relationship in ns2. Anyway, so I wrote those tcl procedures to get the
> ftps to restart.
> I wanted to run more modern TCPs, hence use of the Linux TCP.
>
> The dynamic bandwidth thing changes the bandwidth and preset times. It
> does it by multiplying the original bandwidth so I set up an array of
> the mulipliers. I'm sure there's
> a more elegant way to do it but, at this stage of my life, I have
> accepted that I don't do
> elegant. I'm more of a "get the job done" kind of coder. And I really
> don't like ns2 so I
> generally find it works better to just whack it with a hammer rather
> than to try to do
> something beautifully. Note, as above, that a lot of the stuff in ns2
> just doesn't work.
>
> There is code to build up the topology which shouldn't be that hard to
> figure out.
>
> Kathie
>
> On 2/18/15 11:19 PM, kanu monga wrote:
> > is there anyone who can help me in understanding tcl code of
> codel(given
> > below)
> > This one is quite complex for me.
> > please help if you can.
> > Thanks in advance.
> >
> > ###################################################################
> >
> >
> > # ns codel.tcl f w c {b}Mb s d r
> > # where:
> > #f = # ftps
> > #w = # PackMime connections per second
> > #c = # CBRs
> > #b = bottleneck bandwidth in Mbps
> > #s = filesize for ftp, -1 for infinite
> > #d = dynamic bandwidth, if non-zero, changes (kind of kludgey)
> > #have to set the specific change ratios in this file (below)
> > #r = number of "reverse" ftps
> >
> > set stopTime 300
> > set ns [new Simulator]
> >
> > # These are defaults if values not set on command line
> >
> > set num_ftps 1
> > set web_rate 0
> > set revftp 0
> > set num_cbrs 0
> > #rate and packetSize set in build_cbr
> > set bottleneck 3Mb
> > #for a 10MB ftp
> > set filesize 10000000
> > set dynamic_bw 0
> > set greedy 0
> >
> > # Parse command line
> >
> > if {$argc >= 1} {
> > set num_ftps [lindex $argv 0]
> > if {$argc >= 2} {
> > set web_rate [lindex $argv 1]
> > if {$argc >= 3} {
> > set num_cbrs [lindex $argv 2]
> > if {$argc >= 4} {
> > set bottleneck [lindex $argv 3]
> > if {$argc >= 5} {
> > set filesize [lindex $argv 4]
> > if {$argc >= 6} {
> > set dynamic_bw [lindex $argv 5]
> > if {$argc >= 7} {
> > set revftp [lindex $argv 6]
> > }
> > }
> > }
> > }
> > }
> > }
> > }
> >
> > set bw [bw_parse $bottleneck]
> > if { $revftp >= 1} {
> > set num_revs $revftp
> > } else {
> > set num_revs 0
> > }
> > puts "ftps $num_ftps webrate $web_rate cbrs $num_cbrs bw $bw filesize
> > $filesize reverse $num_revs"
> >
> > # experiment settings
> > set psize 1500
> > if { $bw < 1000000} { set psize 500 }
> > set nominal_rtt [delay_parse 100ms]
> > set accessdly 20
> > set bdelay 10
> > set realrtt [expr 2*(2*$accessdly + $bdelay)]
> > puts "accessdly $accessdly bneckdly $bdelay realrtt $realrtt
> bneckbw $bw"
> >
> > # CoDel values
> > # interval to keep min over
> > set interval [delay_parse 100ms]
> > # target in ms.
> > set target [delay_parse 5ms]
> >
> > global defaultRNG
> > $defaultRNG seed 0
> > ns-random 0
> > #$defaultRNG seed 54321
> > #ns-random 23145
> >
> > # ------- config info is all above this line ----------
> >
> > #bdp in packets, based on the nominal rtt
> > set bdp [expr round($bw*$nominal_rtt/(8*$psize))]
> >
> > Trace set show_tcphdr_ 1
> > set startTime 0.0
> >
> > #TCP parameters - have to set both for FTPs and PackMime
> >
> > Agent/TCP set window_ [expr $bdp*16]
> > Agent/TCP set segsize_ [expr $psize-40]
> > Agent/TCP set packetSize_ [expr $psize-40]
> > Agent/TCP set windowInit_ 4
> > Agent/TCP set segsperack_ 1
> > Agent/TCP set timestamps_ true
> > set delack 0.4
> > Agent/TCP set interval_ $delack
> >
> > Agent/TCP/FullTcp set window_ [expr $bdp*16]
> > Agent/TCP/FullTcp set segsize_ [expr $psize-40]
> > Agent/TCP/FullTcp set packetSize_ [expr $psize-40]
> > Agent/TCP/FullTcp set windowInit_ 4
> > Agent/TCP/FullTcp set segsperack_ 1
> > Agent/TCP/FullTcp set timestamps_ true
> > Agent/TCP/FullTcp set interval_ $delack
> >
> >
> > Agent/TCP/Linux instproc done {} {
> > global ns filesize
> > #this doesn't seem to work, had to hack tcp-linux.cc to do repeat ftps
> > $self set closed_ 0
> > #needs to be delayed by at least .3sec to slow start
> > puts "[$ns now] TCP/Linux proc done called"
> > $ns at [expr [$ns now] + 0.3] "$self send $filesize"
> > }
> >
> > # problem is that idle() in tcp.cc never seems to get called...
> > Application/FTP instproc resume {} {
> > puts "called resume"
> > global filesize
> > $self send $filesize
> > #$ns at [expr [$ns now] + 0.5] "[$self agent] reset"
> > $ns at [expr [$ns now] + 0.5] "[$self agent] send $filesize"
> > }
> >
> > Application/FTP instproc fire {} {
> > global filesize
> > $self instvar maxpkts_
> > set maxpkts_ $filesize
> > [$self agent] set maxpkts_ $filesize
> > $self send $maxpkts_
> > puts "fire() FTP"
> > }
> >
> > #buffersizes
> > set buffersize [expr $bdp]
> > set buffersize1 [expr $bdp*10]
> >
> > Queue/CoDel set target_ $target
> > Queue/CoDel set interval_ $interval
> >
> > #set Flow_id 1
> >
> > proc build_topology { ns } {
> > # nodes n0 and n1 are the server and client side gateways and
> > # the link between them is the congested slow link. n0 -> n1
> > # handles all the server to client traffic.
> > #
> > # if the web_rate is non-zero, node n2 will be the packmime
> server cloud
> > # and node n3 will be the client cloud.
> > #
> > # num_ftps server nodes and client nodes are created for the ftp
> > sessions.
> > # the first client node is n{2+w} and the first server node is
> n{2+f+w}
> > # where 'f' is num_ftps and 'w' is 1 if web_rate>0 and 0
> otherwise.
> > # servers will be even numbered nodes, clients odd
> > # Warning: the numbering here is ridiculously complicated
> >
> > global bw bdelay accessdly buffersize buffersize1 filesize
> node_cnt
> > set node_cnt 2
> >
> > # congested link
> > global n0 n1
> > set n0 [$ns node]
> > set n1 [$ns node]
> > $ns duplex-link $n0 $n1 $bw ${bdelay}ms CoDel
> > $ns duplex-link-op $n0 $n1 orient right
> > $ns duplex-link-op $n0 $n1 queuePos 0.5
> > $ns duplex-link-op $n1 $n0 queuePos 1.5
> > $ns queue-limit $n0 $n1 $buffersize
> > $ns queue-limit $n1 $n0 $buffersize
> > set node_cnt 2
> >
> > #dynamic bandwidth
> > # these are the multipliers for changing bw, times initial set bw
> > # edit these values to get different patterns
> > global stopTime dynamic_bw
> > array names bw_changes
> > set bw_changes(1) 0.1
> > set bw_changes(2) 0.01
> > set bw_changes(3) 0.5
> > set bw_changes(4) 0.01
> > set bw_changes(5) 1.0
> >
> > puts "bottleneck starts at [[[$ns link $n0 $n1] link] set
> > bandwidth_]bps"
> > for {set k 1} {$k <= $dynamic_bw} {incr k 1} {
> > set changeTime [expr $k*$stopTime/($dynamic_bw+1)]
> > set f $bw_changes($k)
> > set newBW [expr $f*$bw]
> > puts "change at $changeTime to [expr $newBW/1000000.]Mbps"
> > $ns at $changeTime "[[$ns link $n0 $n1] link] set bandwidth_ $newBW"
> > $ns at $changeTime "[[$ns link $n1 $n0] link] set bandwidth_ $newBW"
> > $ns at $changeTime "puts $newBW"
> > }
> >
> > set li_10 [[$ns link $n1 $n0] queue]
> > set li_01 [[$ns link $n0 $n1] queue]
> >
> > set tchan_ [open /tmp/redqvar.tr <http://redqvar.tr>
> <http://redqvar.tr> w]
> > $li_01 trace curq_
> > $li_01 trace d_exp_
> > $li_01 attach $tchan_
> >
> > global num_ftps web_rate num_cbrs greedy num_revs
> > set linkbw [expr $bw*10]
> >
> > set w [expr $web_rate > 0]
> > if {$w} {
> > global n2 n3
> > #server
> > set n2 [$ns node]
> > $ns duplex-link $n2 $n0 $linkbw ${accessdly}ms DropTail
> > $ns queue-limit $n2 $n0 $buffersize1
> > $ns queue-limit $n0 $n2 $buffersize1
> >
> > #client
> > set n3 [$ns node]
> > $ns duplex-link $n1 $n3 $linkbw ${accessdly}ms DropTail
> > $ns queue-limit $n1 $n3 $buffersize1
> > $ns queue-limit $n3 $n1 $buffersize1
> > set node_cnt 4
> > }
> > #need to fix the angles if use nam
> > for {set k 0} {$k < $num_ftps} {incr k 1} {
> > # servers
> > set j $node_cnt
> > global n$j
> > set n$j [$ns node]
> > if {$greedy > 0 && $k == 0} {
> > $ns duplex-link [set n$j] $n0 $linkbw 1ms DropTail
> > } else {
> > $ns duplex-link [set n$j] $n0 $linkbw ${accessdly}ms DropTail
> > }
> > $ns queue-limit [set n$j] $n0 $buffersize1
> > $ns queue-limit $n0 [set n$j] $buffersize1
> > set angle [expr $num_ftps>1? 0.75+($k-1)*.5/($num_ftps-1) : 1]
> > $ns duplex-link-op $n0 [set n$j] orient $angle
> > incr node_cnt
> >
> > # clients
> > set j $node_cnt
> > global n$j
> > set n$j [$ns node]
> > set dly [expr ${accessdly} +($k+1)]
> > $ns duplex-link $n1 [set n$j] $linkbw ${dly}ms DropTail
> > $ns queue-limit $n1 [set n$j] $buffersize1
> > $ns queue-limit [set n$j] $n1 $buffersize1
> > set angle [expr $num_ftps>1?
> fmod(2.25-($k-1)*.5/($num_ftps-1),
> > 2) : 0]
> > $ns duplex-link-op $n1 [set n$j] orient $angle
> > incr node_cnt
> > }
> > for {set k 0} {$k < $num_cbrs} {incr k 1} {
> > # servers
> > set j $node_cnt
> > global n$j
> > set n$j [$ns node]
> > $ns duplex-link [set n$j] $n0 $linkbw ${accessdly}ms DropTail
> > $ns queue-limit [set n$j] $n0 $buffersize1
> > $ns queue-limit $n0 [set n$j] $buffersize1
> > # set angle [expr $num_cbrs>1? 0.75+($k-1)*.5/($num_cbrs-1)
> : 1]
> > $ns duplex-link-op $n0 [set n$j] orient $angle
> > incr node_cnt
> >
> > # clients
> > set j $node_cnt
> > global n$j
> > set n$j [$ns node]
> > $ns duplex-link $n1 [set n$j] $linkbw ${accessdly}ms
> DropTail
> > $ns queue-limit $n1 [set n$j] $buffersize1
> > $ns queue-limit [set n$j] $n1 $buffersize1
> > # set angle [expr $num_cbrs>1?
> fmod(2.25-($k-1)*.5/($num_ftps-1),
> > 2) : 0]
> > $ns duplex-link-op $n1 [set n$j] orient $angle
> > incr node_cnt
> > }
> > #reverse direction ftps
> > for {set k 0} {$k < $num_revs} {incr k 1} {
> > # clients
> > set j $node_cnt
> > global n$j
> > set n$j [$ns node]
> > $ns duplex-link [set n$j] $n0 $linkbw ${accessdly}ms DropTail
> > $ns queue-limit [set n$j] $n0 $buffersize1
> > $ns queue-limit $n0 [set n$j] $buffersize1
> > set angle [expr $num_ftps>1? 0.75+($k-1)*.5/($num_ftps-1) : 1]
> > $ns duplex-link-op $n0 [set n$j] orient $angle
> > incr node_cnt
> >
> > # servers
> > set j $node_cnt
> > global n$j
> > set n$j [$ns node]
> > set dly [expr ($accessdly)*1.1 +($k+1)]
> > $ns duplex-link $n1 [set n$j] $linkbw ${dly}ms DropTail
> > $ns queue-limit $n1 [set n$j] $buffersize1
> > $ns queue-limit [set n$j] $n1 $buffersize1
> > set angle [expr $num_ftps>1?
> fmod(2.25-($k-1)*.5/($num_ftps-1),
> > 2) : 0]
> > $ns duplex-link-op $n1 [set n$j] orient $angle
> > incr node_cnt
> > }
> > }
> >
> > proc build_cbr {cnd snd startTime timeToStop Flow_id} {
> > global ns
> > set udp [$ns create-connection UDP $snd LossMonitor $cnd $Flow_id]
> > set cbr [new Application/Traffic/CBR]
> > $cbr attach-agent $udp
> > # change these for different types of CBRs
> > $cbr set packetSize_ 100
> > $cbr set rate_ 0.064Mb
> > $ns at $startTime "$cbr start"
> > $ns at $timeToStop "$cbr stop"
> > }
> >
> > # cnd is client node, snd is server node
> > proc build_ftpclient {cnd snd startTime timeToStop Flow_id} {
> >
> > global ns filesize greedy revftp
> > set ctcp [$ns create-connection TCP/Linux $snd TCPSink/Sack1 $cnd
> > $Flow_id]
> > $ctcp select_ca cubic
> > set ftp [$ctcp attach-app FTP]
> > $ftp set enableResume_ true
> > $ftp set type_ FTP
> >
> > #set up a single infinite ftp with smallest RTT
> > if {$greedy > 0 || $filesize < 0} {
> > $ns at $startTime "$ftp start"
> > set greedy 0
> > } else {
> > $ns at $startTime "$ftp send $filesize"
> > }
> > $ns at $timeToStop "$ftp stop"
> > }
> >
> > proc build_webs {cnd snd rate startTime timeToStop} {
> > set CLIENT 0
> > set SERVER 1
> >
> > # SETUP PACKMIME
> > set pm [new PackMimeHTTP]
> > $pm set-TCP Sack
> > $pm set-client $cnd
> > $pm set-server $snd
> > $pm set-rate $rate; # new connections per
> second
> > $pm set-http-1.1; # use HTTP/1.1
> >
> > # create RandomVariables
> > set flow_arrive [new RandomVariable/PackMimeHTTPFlowArrive $rate]
> > set req_size [new RandomVariable/PackMimeHTTPFileSize $rate
> $CLIENT]
> > set rsp_size [new RandomVariable/PackMimeHTTPFileSize $rate
> $SERVER]
> >
> > # assign RNGs to RandomVariables
> > $flow_arrive use-rng [new RNG]
> > $req_size use-rng [new RNG]
> > $rsp_size use-rng [new RNG]
> >
> > # set PackMime variables
> > $pm set-flow_arrive $flow_arrive
> > $pm set-req_size $req_size
> > $pm set-rsp_size $rsp_size
> >
> > global ns
> > $ns at $startTime "$pm start"
> > $ns at $timeToStop "$pm stop"
> > }
> >
> > proc uniform {a b} {
> > expr $a + (($b- $a) * ([ns-random]*1.0/0x7fffffff))
> > }
> >
> > proc finish {} {
> > global ns
> > $ns halt
> > $ns flush-trace
> > exit 0
> > }
> >
> > # $ns namtrace-all [open out.nam w]
> > # $ns color 2 blue
> > # $ns color 3 red
> > # $ns color 4 yellow
> > # $ns color 5 green
> >
> > build_topology $ns
> >
> > #$ns trace-queue $n0 $n1 [open out_n0ton1.tr
> <http://out_n0ton1.tr> <http://out_n0ton1.tr> w]
> > #set fname f${num_ftps}w${web_rate}b${bottleneck}.tr
> > set fname f.tr <http://f.tr> <http://f.tr>
> > puts $fname
> > $ns trace-queue $n0 $n1 [open /tmp/$fname w]
> > #reverse direction
> > #$ns trace-queue $n1 $n0 [open /tmp/$fname w]
> >
> > set node_cnt 2
> > if {$web_rate > 0} {
> > build_webs $n3 $n2 $web_rate 0 $stopTime
> > set node_cnt 4
> > }
> >
> > for {set k 1} {$k <= $num_ftps} {incr k 1} {
> > set j $node_cnt
> > incr node_cnt
> > set i $node_cnt
> > build_ftpclient [set n$i] [set n$j] \
> > $startTime $stopTime $i
> > # [expr 1.0*($k-1)] $stopTime $i
> > # [expr $startTime+($k-1)*[uniform 0.0 2.0]] $stopTime $i
> > incr node_cnt
> > }
> >
> > for {set k 1} {$k <= $num_cbrs} {incr k 1} {
> > set j $node_cnt
> > incr node_cnt
> > set i $node_cnt
> > build_cbr [set n$i] [set n$j] \
> > [expr $startTime+($k-1)*[uniform 0.0 2.0]] $stopTime $i
> > incr node_cnt
> > }
> >
> > #for reverse direction, give client smaller number
> > for {set k 1} {$k <= $num_revs} {incr k 1} {
> > set j $node_cnt
> > incr node_cnt
> > set i $node_cnt
> > build_ftpclient [set n$j] [set n$i] $startTime $stopTime $j
> > incr node_cnt
> > }
> >
> > $ns at [expr $stopTime ] "finish"
> > $ns run
> > exit 0
> > ###################################################################
> >
> >
> > _______________________________________________
> > Codel mailing list
> > Codel at lists.bufferbloat.net <mailto:Codel at lists.bufferbloat.net>
> > https://lists.bufferbloat.net/listinfo/codel
> >
>
>
>
> ------------------------------
>
> Message: 2
> Date: Thu, 19 Feb 2015 13:38:53 -0800
> From: Dave Taht <dave.taht at gmail.com <mailto:dave.taht at gmail.com>>
> To: Kathleen Nichols <nichols at pollere.com <mailto:nichols at pollere.com>>
> Cc: "codel at lists.bufferbloat.net
> <mailto:codel at lists.bufferbloat.net>" <codel at lists.bufferbloat.net
> <mailto:codel at lists.bufferbloat.net>>
> Subject: Re: [Codel] Codel code
> Message-ID:
>
> <CAA93jw5dWzXfEX70w1PFmPudH+qfurq=e4Srbth2+f4QC2h4Og at mail.gmail.com
> <mailto:e4Srbth2%2Bf4QC2h4Og at mail.gmail.com>>
> Content-Type: text/plain; charset=UTF-8
>
> I am fairly happy with the example codel code that is now in the ns3
> branch,
> which is pure c++, and quite a bit easier to understand than tcl, to
> my eye.
>
> https://www.nsnam.org/doxygen/codel-vs-droptail-asymmetric_8cc.html
>
> regrettably the fq_codel version for ns3 is still out of tree.
>
>
> ------------------------------
>
> Message: 3
> Date: Fri, 20 Feb 2015 00:46:11 -0800
> From: Anjali Chawla <anjaliachawla013 at gmail.com
> <mailto:anjaliachawla013 at gmail.com>>
> To: codel at lists.bufferbloat.net <mailto:codel at lists.bufferbloat.net>
> Subject: [Codel] Setpoint in CoDel
> Message-ID:
>
> <CAL6QuG8ktPQgrGkA+wp9V=BHqdcQh4E19dsX02Pozh_fjfJ+0w at mail.gmail.com
> <mailto:BHqdcQh4E19dsX02Pozh_fjfJ%2B0w at mail.gmail.com>>
> Content-Type: text/plain; charset="utf-8"
>
> Hi
> anyone please explain me what is concept behind setpoint in codel
> algorithm??
> Thanks
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL:
> <https://lists.bufferbloat.net/pipermail/codel/attachments/20150220/285d07dd/attachment-0001.html>
>
> ------------------------------
>
> Message: 4
> Date: Fri, 20 Feb 2015 01:21:13 -0800
> From: Dave Taht <dave.taht at gmail.com <mailto:dave.taht at gmail.com>>
> To: Anjali Chawla <anjaliachawla013 at gmail.com
> <mailto:anjaliachawla013 at gmail.com>>
> Cc: "codel at lists.bufferbloat.net
> <mailto:codel at lists.bufferbloat.net>" <codel at lists.bufferbloat.net
> <mailto:codel at lists.bufferbloat.net>>
> Subject: Re: [Codel] Setpoint in CoDel
> Message-ID:
>
> <CAA93jw43TJdo9FK61HK0et3Mopq8-YKjjkNm_6_OF7+g0iMipw at mail.gmail.com
> <mailto:CAA93jw43TJdo9FK61HK0et3Mopq8-YKjjkNm_6_OF7%2Bg0iMipw at mail.gmail.com>>
> Content-Type: text/plain; charset=UTF-8
>
> the best talk on the subject was van?s at ietf, it is well worth
> watching multiple times.
>
> http://www.bufferbloat.net/projects/cerowrt/wiki/Bloat-videos
>
> He goes deeply into the reasoning behind the sojourn time, and setpoint.
>
> About the only other somewhat in-depth talk on how codel really works
> was my attempt at stanford.
>
> and there are of course other talks above that try to describe bits of
> the elephant.
>
> One of these days I hope someone puts together the definitive talk on
> codel complete with even more elegant demos than stephen hemminger
> came up with..., but it does seem to be really hard for people to get
> at a deep level... most of my own talks are more about fq than
> fq_codel, and that is *tons* easier to explain.
>
> If you watch them all a bunch of times, and run a few experiments of
> your own, enlightenment will come. For me, it was seeing kathie give
> the first public explanation of codel - about midway through I started
> vibrating in my seat - but I had just completed a slash and burn tour
> through all the aqm literature going back 30 years, and trying
> everything that seemed to make sense. I don?t know how to impart that
> in an email...
>
>
> On Fri, Feb 20, 2015 at 12:46 AM, Anjali Chawla
> <anjaliachawla013 at gmail.com <mailto:anjaliachawla013 at gmail.com>> wrote:
> > Hi
> > anyone please explain me what is concept behind setpoint in codel
> > algorithm??
> > Thanks
> >
> > _______________________________________________
> > Codel mailing list
> > Codel at lists.bufferbloat.net <mailto:Codel at lists.bufferbloat.net>
> > https://lists.bufferbloat.net/listinfo/codel
> >
>
>
>
> --
> Dave T?ht
>
> thttp://www.bufferbloat.net/projects/bloat/wiki/Upcoming_Talks
> <http://www.bufferbloat.net/projects/bloat/wiki/Upcoming_Talks>
>
>
> ------------------------------
>
> Message: 5
> Date: Fri, 20 Feb 2015 10:01:54 -0800
> From: Dave Taht <dave.taht at gmail.com <mailto:dave.taht at gmail.com>>
> To: Anjali Chawla <anjaliachawla013 at gmail.com
> <mailto:anjaliachawla013 at gmail.com>>
> Cc: "codel at lists.bufferbloat.net
> <mailto:codel at lists.bufferbloat.net>" <codel at lists.bufferbloat.net
> <mailto:codel at lists.bufferbloat.net>>
> Subject: Re: [Codel] Setpoint in CoDel
> Message-ID:
>
> <CAA93jw7147Ucy=fRmt-65VuUSDyVwHwH9NCmOFjq4GULYGC63A at mail.gmail.com
> <mailto:fRmt-65VuUSDyVwHwH9NCmOFjq4GULYGC63A at mail.gmail.com>>
> Content-Type: text/plain; charset=UTF-8
>
> On Fri, Feb 20, 2015 at 1:21 AM, Dave Taht <dave.taht at gmail.com
> <mailto:dave.taht at gmail.com>> wrote:
> > the best talk on the subject was van?s at ietf, it is well worth
> > watching multiple times.
> >
> > http://www.bufferbloat.net/projects/cerowrt/wiki/Bloat-videos
> >
> > He goes deeply into the reasoning behind the sojourn time, and
> setpoint.
> >
> > About the only other somewhat in-depth talk on how codel really works
> > was my attempt at stanford.
>
> The stanford talk took some digging to find. I would really like a shot
> at a do-over one day - or for van to give his talk again to be filmed at
> high quality.
>
> https://www.youtube.com/watch?v=Mxoa5Si4Ubw
>
> >
> > and there are of course other talks above that try to describe bits of
> > the elephant.
> >
> > One of these days I hope someone puts together the definitive talk on
> > codel complete with even more elegant demos than stephen hemminger
> > came up with..., but it does seem to be really hard for people to get
> > at a deep level... most of my own talks are more about fq than
> > fq_codel, and that is *tons* easier to explain.
> >
> > If you watch them all a bunch of times, and run a few experiments of
> > your own, enlightenment will come. For me, it was seeing kathie give
> > the first public explanation of codel - about midway through I started
> > vibrating in my seat - but I had just completed a slash and burn tour
> > through all the aqm literature going back 30 years, and trying
> > everything that seemed to make sense. I don?t know how to impart that
> > in an email...
> >
> >
> > On Fri, Feb 20, 2015 at 12:46 AM, Anjali Chawla
> > <anjaliachawla013 at gmail.com <mailto:anjaliachawla013 at gmail.com>>
> wrote:
> >> Hi
> >> anyone please explain me what is concept behind setpoint in codel
> >> algorithm??
> >> Thanks
> >>
> >> _______________________________________________
> >> Codel mailing list
> >> Codel at lists.bufferbloat.net <mailto:Codel at lists.bufferbloat.net>
> >> https://lists.bufferbloat.net/listinfo/codel
> >>
> >
> >
> >
> > --
> > Dave T?ht
> >
> > thttp://www.bufferbloat.net/projects/bloat/wiki/Upcoming_Talks
> <http://www.bufferbloat.net/projects/bloat/wiki/Upcoming_Talks>
>
>
>
> --
> Dave T?ht
>
> thttp://www.bufferbloat.net/projects/bloat/wiki/Upcoming_Talks
> <http://www.bufferbloat.net/projects/bloat/wiki/Upcoming_Talks>
>
>
> ------------------------------
>
> _______________________________________________
> Codel mailing list
> Codel at lists.bufferbloat.net <mailto:Codel at lists.bufferbloat.net>
> https://lists.bufferbloat.net/listinfo/codel
>
>
> End of Codel Digest, Vol 29, Issue 4
> ************************************
>
>
>
>
> _______________________________________________
> Codel mailing list
> Codel at lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/codel
>
More information about the Codel
mailing list