Re: [RFC/PATCH] IMQ port to 2.6

From: jamal (hadi_at_cyberus.ca)
Date: 01/26/04

  • Next message: Marcel Holtmann: "Re: Bluetooth USB oopses on unplug (2.6.1)" 
    To: "Vladimir B. Savkin" <master@sectorb.msk.ru>
Date:	26 Jan 2004 08:38:33 -0500

    On Mon, 2004-01-26 at 04:32, Vladimir B. Savkin wrote:

    > On Sun, Jan 25, 2004 at 10:09:48PM -0500, jamal wrote:
    [..]
    > > shape). I have not seen anything in favor of shaping; i could be wrong
    > > (so if you know of something or have experimented pass the data).
    >
    > Yes, I have experimented. Shaping works much better:
    > much less packets dropped, much better donwload rates for clients.
    >

    I cant say i doubt you, but your word alone is insufficient data ;->

    The important point is the eventual effective throughput and fairness
    amongst the flows. Whether it is induced by an increased RTT from
    shaping or a single packet retransmit on some misbehaving flows because
    of policing is less important. i.e it is not evil for packets to
    be dropped.
    When you analyse something like this you should look at the aggregate
    throughput instead of a single client with better downloads (probably at
    the expense of another poor client download).

    > I want to shape traffic that comes from upstream to clients connected
    > via PPTP.

    So if i understand correctly and was to draw this:
    you have clients on the left side coming in through ethx and that need
    to be tunneled to some pppoe/pptp before going out ethy on the right
    hand side. The right handside represents "upstream" in your terminology.
    Is this correct? I hate it when people ask me for a diagram for
    something that looks obvious;-> but bear with me and supply me with a
    diagram if i didnt understand you.

    >
    > Here is a part of my scripts:
    >
    > DEVICE=imq0
    > /sbin/tc qidisc add dev $DEVICE root handle 10: htb r2q 1 default 100
    > /sbin/tc class add dev $DEVICE parent 10:0 classid 10:1 est 1sec 8sec htb \
    > rate 10Mbit burst 400k
    > /sbin/tc class add dev $DEVICE parent 10:1 classid 10:2 est 1sec 8sec htb \
    > rate 180kbps ceil 180kbps burst 3000
    > # default class for users
    > /sbin/tc class add dev $DEVICE parent 10:2 classid 10:101 est 1sec 8sec htb \
    > rate 20kbps burst 1k ceil 50kbps cburst 1k
    > /sbin/tc qdisc add dev $DEVICE parent 10:101 wrr \
    > dest ip 128 1 wmode1=1 wmode2=1
    > /sbin/tc filter add dev $DEVICE protocol ip parent 10:0 \
    > prio 100 handle 1 fw flowid 10:101
    > # more classes to follow ...
    >

    So why not have the above attached to ethy? Why does it have to be done
    at some other device?

    >
    > The limit 50kbps is artificial, so there's no bottleneck in
    > connection from upstream to this router. I cannot allocate all
    > the channel bandwidth to clients for some political reasons.
    > Then, I mark packets I want to go to this default user class with mark "1",
    > like this:
    >
    > iptables -t mangle -A FORWARD -i $UPLINK_DEV -d $CLIENTS_NET \
    > -j IMQ --todev 0 # traffic from internet to clients
    > iptables -t mangle -A FORWARD -i $UPLINK_DEV -d $CLIENTS_NET \
    > -j MARK --set-mark 1 # default class

    Why do you need the redirect to IMQ?
    If you can selectively mark packets here (or at any other netfilter
    hook) you could use the fwmark classifier to attach to different
    10:x classes on the ethy interface. I feel i am missing something.

    > So, I shape traffic destined to clients, and I use "wrr" to
    > divide bandwidth fairly. I cannot attach qdisc to an egress device
    > because there's no single one, each client has its own ppp interface.
    >

    I mean the ethy interface not the ppp* interfaces. Mark the packets;
    use fwmark classifier.

    > Well, I could move this shaping upstream, but what if upstream router was
    > some dumb cisco with no "wrr" qdisc?

    You dont have to.
    Give me the diagram.

    cheers,
    jamal

    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.kernel.org
    More majordomo info at http://vger.kernel.org/majordomo-info.html
    Please read the FAQ at http://www.tux.org/lkml/

  • Next message: Marcel Holtmann: "Re: Bluetooth USB oopses on unplug (2.6.1)"

    Relevant Pages

    • Re: Multicast or not
      ... AliR wrote: ... I have one server application which needs to talked to ... > server application needs to send the exact same packets to all the clients, ... > events and the threads would send the packets. ...
      (microsoft.public.vc.mfc)
    • Migrating from Word mail merge to Access Reports
      ... I'm trying to help convert a fairly cumbersome MS WordMS Access mail merge ... to the clients' own info via a typical mail merge. ... list and then producing however many packets are ... the 22" vertical size limit for an Access report. ...
      (microsoft.public.access.reports)
    • Solaris 9 connection problems
      ... The server is behind a firewall which is not under ... my control although I am assured that the subnet that the test clients ... On the server I see the Syn packets arriving and Syn Ack and Ack packets ... I'm also confused why this is affecting connections from some clients ...
      (SunManagers)
    • Re: [RFC/PATCH] IMQ port to 2.6
      ... I have not seen anything in favor of shaping; ... much less packets dropped, much better donwload rates for clients. ... I want to shape traffic that comes from upstream to clients connected ...
      (Linux-Kernel)
    • Re: [RFC/PATCH] IMQ port to 2.6
      ... >> much less packets dropped, much better donwload rates for clients. ... > shaping or a single packet retransmit on some misbehaving flows because ... >> connection from upstream to this router. ...
      (Linux-Kernel)