From: Carlos G Mendioroz (tron@huapi.ba.ar)
Date: Sun May 23 2004 - 19:29:40 GMT-3
Tom,
there is more than one kind of policer out there...
Some precisions in the way we use the word congested/congestion would
help understand.
Some of the tools/mechs of CBWFQ only engage when the tx-ring is full
(that's the hardware FIFO output queue attached to each interface).
If we use "congested" to define this scenario, then all your CBWFQ
bandwidth and priority parameters are only used when the interface is
congested.
You can go over your priority assigned bandwidth if not congested (and
there is an implicit policer with every priority queue).
Now, if you use an explicit policer or shaper, then that is always
active, interface being congested or not.
Both shaping and policing do not let traffic go over your parametric
quota (the plateau you were talking before, which may be also like an
elephant inside a snake :-).
Shapers do that by delaying exceeding traffic, policers by dropping.
(Well, if you are marking, you will be able to exceed, but only with
marked traffic).
Sounds logical ?
Richard Dumoulin wrote:
> No, its the other way around. If the actual speed is 10Mbps and you're
> shaping at 96 Kbps then you are creating congestion at 96Kbps because the
> packets will be queued once the rate reaches 96Kbps,
>
> --Richard
>
> -----Original Message-----
> From: Tom Rogers [mailto:cccie71@yahoo.com]
> Sent: domingo, 23 de mayo de 2004 3:27
> To: Richard Dumoulin; Ahmed Mustafa
> Cc: ccielab@groupstudy.com
> Subject: RE: QoS Shape Vs Police
>
>
> Richard and Ahmed,
>
> So I take that shape is a congestion, command right ?
>
> If so, that means if I I have shape average 96000 ===> Only during
> congestion class XXX is shaped to 96 kbps and if there is no congestion I ll
> send full e0 speed out, right?
>
> Rx(e0)------------------------(e0)ISP
> policy-map TEST
> class xxx
> shape 96000
> int e0/0
> service-policy out TEST
>
> Thats what I thought. So if ISP only wants 96kbps to come in, should'nt we
> be using police keyword instead of shape on Rx ?
>
> But the DQoS book is using shape keyword (ISBN#1-58720-058-9, page 364, Eg
> 5-5)
> It says Provider is policing at 96kbps and adjust your traffic accordingly.
>
> Thanx
> Tom
>
> Richard Dumoulin <richard.dumoulin@vanco.es> wrote:
>
> Tom, it is precisely traffic shaping that creates congestion. When the
> packet rate approach the CIR then they are enqueued in the pvc queues and
> shaped.
>
>
> --Richard
>
> -----Original Message-----
> From: Tom Rogers [mailto:cccie71@yahoo.com <mailto:cccie71@yahoo.com> ]
> Sent: domingo, 23 de mayo de 2004 0:40
> To: Ahmed Mustafa
> Cc: ccielab@groupstudy.com
> Subject: Re: QoS Shape Vs Police
>
>
> Ahmed,
> Thank you for your explanation. My confision is still not clear. I
> understand that in shape the graph is like a plateau ie packets re held as
> long long as they can and transmitted at a certain threshold say 96kbps in
> this book example.
>
>
> What I want to know is, is this only during the CONGESTION ?
> Or no matter what, whether congestion or not... like in policing.
>
> Is this shape independendent of congestion ?
>
> Thanx in advance
>
> Tom
>
> Ahmed Mustafa <ahmed.mustafa@sbcglobal.net> wrote:
> Tom,
>
> The major difference between policing and shaping is when you do policing,
> no matter if your link is 10MB, 100MB, 1GB, once you set the limit, and
> traffic beyond that limit will be dropped or you set parameters if you don't
> want the traffic to get dropped. In shapping, your intention is not to drop
> the packet, but to delay the packet so the packet doesn't get dropped
>
> Shapping is advantageous where your low end traffic will not get dropped,
> and it will arrive to the destination with some latency. Shapping is never
> meant to be for delay sensitive traffic such as voice and video.
>
> The best example would be
>
> Your router is connected to ISP. The Clock rate of your line is 1536, but
> the CIR you are paying for is 128KB. ISP knows that the guranteed CIR is
> 128KB, but the router at your end could easily send traffic above 128KB. The
> ISP router can then start policing traffic coming from your end at 128KB at
> the ingress point, so no matter how fast your access rate the traffic above
> 128 will get dropped.
>
> So now at your end you could configure some shapping at the egress point
> where you can shape your data traffic at certain rate, and prioritized your
> low latency traffic at certain rate.
>
> HTH,
>
> Ahmed
> ----- Original Message -----
> From: "Tom Rogers"
> To:
> Sent: Saturday, May 22, 2004 7:55 AM
> Subject: QoS Shape Vs Police
>
>
>
>>Hello Every one,
>>
>>Can some explain me the difference beween shape and police.
>>I did the research and could not find an answer.
>>
>>When I look at the DQoS book ISBN#1-58720-058-9
>>Page 364 Example 5-5
>>It says the FR provider wants to police the VCs at 96kbps and we re to
>
> adjust our FR interfaces accordingly.
>
>>I was under the impression police cir command should do it. But the
>>solution has shape average.
>>
>>My confusion is, I thought shape average kicks in when there is a
>
> congestion and would plateau the graph and policing no matter what we limit
> or restrict traffic even if there is no congestion and would tail drop the
> graph.
>
>
>>Can some one tell me the when to use police and when use shape?
>>
>>I think I am lacking some concept here
>>
>>
>>---------------------------------
>>Do you Yahoo!?
>>Yahoo! Domains - Claim yours for only $14.70/year
>>
>>______________________________________________________________________
>>_
>>Please help support GroupStudy by purchasing your study materials from:
>>http://shop.groupstudy.com <http://shop.groupstudy.com/>
>>
>>Subscription information may be found at:
>>http://www.groupstudy.com/list/CCIELab.html
>
> <http://www.groupstudy.com/list/CCIELab.html>
>
> _______________________________________________________________________
> Please help support GroupStudy by purchasing your study materials from:
> http://shop.groupstudy.com <http://shop.groupstudy.com/>
>
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
> <http://www.groupstudy.com/list/CCIELab.html>
>
> ---------------------------------
> Do you Yahoo!?
> Yahoo! Domains - Claim yours for only $14.70/year
>
> _______________________________________________________________________
> Please help support GroupStudy by purchasing your study materials from:
> http://shop.groupstudy.com <http://shop.groupstudy.com/>
>
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
> <http://www.groupstudy.com/list/CCIELab.html>
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam? Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
>
> _______________________________________________________________________
> Please help support GroupStudy by purchasing your study materials from:
> http://shop.groupstudy.com
>
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
>
-- Carlos G Mendioroz <tron@huapi.ba.ar> LW7 EQI Argentina
This archive was generated by hypermail 2.1.4 : Wed Jun 02 2004 - 11:12:15 GMT-3