From: Roger RPF (rpf@bluemail.ch)
Date: Wed Oct 29 2008 - 05:11:54 ARST
Hey Guys, thank you all for the answers.
Sure, that's the solution. I'll use shaped mode on the CPE and play with the
values, that they match my policer values on the Access Switch side.
This should work and with this I can keep it simple on the CPE side. Note,
this might be a strange case, and most of the time I would use CBWFQ but
this wouldn't match the service description for the customer....
Thanks again
Roger
-----Urspr|ngliche Nachricht-----
Von: Daniel Kutchin [mailto:daniel@kutchin.com]
Gesendet: Dienstag, 28. Oktober 2008 18:51
An: 'Roger RPF'
Betreff: RE: AW: question regarding queueing on Catalyst
Roger ---
One more thing to note: unlike shared weights, SHAPED weights
are ABSOLUTE weights.
ABSOLUTE means you use shape_wt to directly calculate
Q-BW (= 1/shaped_wt * interface_bw )
Borrowing an example from Petr's blog to demonstrate what
BW will be allocated to each Q[ueue]:
http://blog.internetworkexpert.com/2008/06/26/quick-notes-on-the-3560-egress
-queuing/#more-141
mls qos
!
interface FastEthernet0/13
switchport mode access
load-interval 30
speed 10 <--- interface BW
duplex full <--- I added this one
srr-queue bandwidth share 33 33 33 1
srr-queue bandwidth shape 50 0 0 0
srr-queue bandwidth limit 20 <--- Affects only shared_BW
Q1
shaped wt = 50, Shared wt = 33
==> Shared wt is ignored because shaped wt is non-zero.
==> BW = 1/50 * (10Mbps) (Note that "srr-q band lim 20" is also ignored)
==> BW = 200 kbps
Q2-Q4
Shaped_wt =0
shared_wts are Q2:33,Q3:33 and Q4:1
srr-queue bandwidth limit 20 ==> 20% of 10Mbps ==> 2Mbps
==> remaining output BW to share = 2 - 0.2 = 1.8Mbps
==> Output BW (Q2) = 33/(33+33+1) * 1.8 = 887kbps
==> Output BW (Q3) = 33/(33+33+1) * 1.8 = 887kbps
==> Output BW (Q4) = 1/(33+33+1) * 1.8 = 26kbps
OK, if you want to police and don't want to use MQC, then you could consider
the storm-control command on the other switch (it act on ingress traffic)
---Daniel
-----Original Message----- From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of Bob Sinclair Sent: Dienstag, 28. Oktober 2008 16:18 To: Roger RPF Cc: reza@toghraee.com; 'GS CCIE-Lab' Subject: Re: AW: question regarding queueing on Catalyst
Hi Roger,
It sounds to me like you might want to consider using the srr-queue bandwidth shape approach outbound on the CPE switch. This would guarantee, limit and queue the excess. Compared to policing on the access switch which drops the excess, shaping would queue the excess.
-HTH,
-Bob
Roger RPF wrote:
Hi Bob, Reza, Thanks a lot for your answer. So, ok, I could get the desired result with setting the port speed, but then I can only set them to 10Mb / 100MB or even 1000MB. Second option, using srr-queue bandwidth limit, this would cause rate-limiting after the specified limit is reached if I understand that right, isn't it? http://www.cisco.com/en/US/docs/switches/lan/catalyst3560/software/release/1 2.2_44_se/configuration/guide/swqos.html#wp1168422 Assume the following. I have a CPE connected to a Access-Swith. On the access-switch I do an aggregate policing to the value which the customer bought. Let's assume, the customer bought 5Mb of overall traffic. On my CPE (still owned by the SP), I have a FastEthernet connecting to the Access-Switch and I do srr-queueing in outgoing direction towards the SP (shared mode). So now, the queueing will never kick in because in regards to the interface on the CPE, it is never congested (interface is 100Mb). But since I have policing on the Access-Switch set to 5Mb, once the Customer traffic reaches 5Mb, it will start dropping. I would like that the CPE starts queueing according to the "policing values" which I've set on the Access Switch", without configuring CBWFQ. I was thinking that this should somehow be possible with using the BW statement or similar... Maybe it's a strange case but is that somehow possible? regards Roger -----Urspr|ngliche Nachricht----- Von: Bob Sinclair [ mailto:bob@bobsinclair.net ] Gesendet: Dienstag, 28. Oktober 2008 15:24 An: Roger RPF Cc: 'GS CCIE-Lab' Betreff: Re: question regarding queueing on Catalyst Hi Roger, In my experience you cannot induce congestion on this switch by changing the software reference bandwidth on the interface. Instead, try changing the port speed, or do an "srr-queue bandwidth limit" command. Remember that the configured values for the srr-queue bandwidth share command are not literally percentages, but relative weights. They appear as percentages only when they happen to add to 100. You might want to check the command reference on this one. Also keep in mind that shaped weights override shared weights. By default queue 1 has a shaped weight of 25, so the "40" weight in your example would not be used unless you zero-out the default shaped weight for queue 1. With the default shaping weights, queues 2, 3, and 4 would share remaining bandwidth in the ratio 10/60, 20/60 and 30/60, respectively. HTH, -Bob Sinclair CCIE 10427 CCSI 30427 www.netmasterclass.net Roger RPF wrote:
Hi Group, I have a question regarding queueing on the switches, for example on a
3560.
If I have a FastEthernet port and I enable srr-queueing as below: mls qos Int Fa0/15 srr-queue bandwidth share 40 10 20 30 The configured values, reflect the percentage which the queue in case of congestion is allocated. Which bandwidth value does the switch take to calculate the real values? The Interface bandwidth, in our case 100Mb?
If
I configure for example bandwidth 10, are the calculations based on 10Mb?
So
the switch will start already start queueing (think that he is concested) when traffic reaches 10Mb? many thanks in advance Roger
Blogs and organic groups at http://www.ccie.net _______________________________________________________________________ Subscription information may be found at: http://www.groupstudy.com/list/CCIELab.html
Blogs and organic groups at http://www.ccie.net _______________________________________________________________________ Subscription information may be found at: http://www.groupstudy.com/list/CCIELab.html
Blogs and organic groups at http://www.ccie.net
This archive was generated by hypermail 2.1.4 : Sat Nov 01 2008 - 15:35:23 ARST