From: Tarun Pahuja (pahujat@gmail.com)
Date: Sat Dec 01 2007 - 11:44:45 ART
Keith,
Scott brought up a very good point. Policing drops the
packets whereas Shaping tries to buffer the packets during the time of
congestion. Rate-limiting works well with controlling Denial of
service attacks and other harmful applications. There isn't a good use
for Rate-limiting as it drops a lot of packets, which makes TCP think
there is congestion, so it slows down the flow and after a while the
flow picks up again and then there is more packet loss. Traffic
shaping on the other hand, just slows the packets down so TCP adapts
to the available bandwidth.
The following statistics were collected from a well documented test.
During the test an ftp application was used to file transfer on a
128Kbps link. When rate limiting was used the performance was only 84
Kpbs as appose to traffic-shaping where the performance was 121 Kps.
Traffic shaping has a limitation that it can not be configured
inbound. Therefore it is recommended to configure traffic shaping
outbound at the customer end when conning to a Service provider using
rate-limiting to enforce CIR. Better performance will be achieved.
HTH,
Tarun
On 12/1/07, Scott Vermillion <scott_ccie_list@it-ag.com> wrote:
> Hey Keith,
>
> I can't think of an article, per se, but Odom's QoS Guide is pretty much
> required reading for CCIE candidates, IMHO. You will not have any doubts
> about this or other such issues if you give that a good read.
>
> In short, just remember that shaping actually delays (buffers) exceeding
> traffic in an attempt to "spool" it out at the CIR. Policing does not
> perform any kind of delaying/buffering, but rather drops or remarks traffic
> that is out of conformance with the defined service parameters. For this
> reason, shaping is typically done outbound while policing is typically done
> inbound. You would shape towards your carrier, for example, because they
> would be policing inbound from you. Better to buffer up your outbound
> traffic and try to spool it out during an ebb in offered load than to have
> the carrier deal more harshly with it inbound. As far as lab scenarios go,
> there are usually keywords in the task which help you to know which to
> choose.
>
> But again, you will not be disappointed with your investment of both time
> and $$ if you buy and read that book. I'm going to read it cover-to-cover
> once more before I attempt the lab in Feb...
>
> Regards,
>
> Scott
>
>
> -----Original Message-----
> From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
> keith tokash
> Sent: Friday, November 30, 2007 9:58 PM
> To: ccielab@groupstudy.com
> Subject: "shape average" vs police
>
> Does anyone happen to know of a nice article explaining the difference
> between
> "shape average" and "police"? I'd like to clear this up, and my searches
> keep
> turning up configuration steps, which are far easier than understanding the
> nuances of the two.
>
> As I understand things, shaping allows some ebb and flow from the average
> rate
> you give it, queueing packets if possible, whereas policing is a hard drop
> at
> the prescribed limit.
>
> With a few exceptions, secrecy is deeply incompatible with democracy and
> with
> science.
> --Carl Sagan
> _________________________________________________________________
> Your smile counts. The more smiles you share, the more we donate. Join in.
> www.windowslive.com/smile?ocid=TXT_TAGLM_Wave2_oprsmilewlhmtagline
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
This archive was generated by hypermail 2.1.4 : Tue Jan 01 2008 - 12:04:28 ARST