From: Petr Lapukhov (petr@internetworkexpert.com)
Date: Fri Sep 26 2008 - 12:09:29 ART
Brian,
As a matter of fact, your priority queue may claim ALL available interface
bandwidth, up to maximum10Mbs limit :) The Catalyst switches do not limit
sending rate of the priority queue - you need to do that yourself, by using
ingress policing or controlling traffic sources. You can only control the
sending rate of non-priority queues, using the SRR shaped mode. So in this
case, your queue gets expedite services, but does not have any rate limits.
You may look at the following post as well:
http://blog.internetworkexpert.com/2008/06/26/quick-notes-on-the-3560-egress-queuing/
for more information on the SRR features.
HTH
-- Petr Lapukhov, CCIE #16379 (R&S/Security/SP/Voice) petr@internetworkexpert.comInternetwork Expert, Inc. http://www.InternetworkExpert.com Toll Free: 877-224-8987 Outside US: 775-826-4344
2008/9/26, Brian Landers <brian@bluecoat93.org>: > > Thanks for the quick reply, Petr! So does that mean if I have the > following > config: > > mls qos > int fa0/4 > speed 100 > srr-queue bandwidth limit 10 > priority-queue out > ! > > ... my queue 1 traffic will be guaranteed 2.5mb of bandwidth (25%, the > default)? > > > On Fri, Sep 26, 2008 at 10:52 AM, Petr Lapukhov < > petr@internetworkexpert.com > > wrote: > > > Hi, > > > > On the 3560 platfor, you can use the command "srr-queue bandwidth limit" > to > > enforce interface-wide shaping (it actually is shaping, since it delays > > packets). Combined with speed manipulation (10/100/1000) you can "shape" > > interface to as little as 1Mbps (with some granularity for speed values, > of > > course). > > > > At the same time, you can configure the use of SRR queues (either in > shaped > > or shared modes) with this feature. In effect, this allows implementing > > hierarchical shaping, with per-interface shaped rate combined with > per-queue > > shaping. > > > > Of course, there are just four-queues and therefore four aggregate > traffic > > classes. But this is the price you have to pay for the high switching > > performance! :) > > > > HTH > > > > > > 2008/9/26, Brian Landers <brian@bluecoat93.org>: > >> > >> Hi all, > >> > >> I have two 3560's connected by Metro Ethernet. The MetroE is policed to > >> 10mb by the service provider, but delivered as 100/full on > copper. Thus, > >> as > >> far as my switches know, they're connected by a 100mb link. > >> > >> Here's the tricky part: I need to do some fun QoS stuff like priority > >> queuing, WRD, and class-based fair-queuing across this link, based on > the > >> 10mb CIR. Currently, since the hardware queues never fill up, the > queuing > >> and WRD never kick in. Is there a way to force the 3560 to use a > >> different > >> value for its "start queueing now" water mark? > >> > >> Yes, this is technically off-topic for the CCIE list, but it seems like > >> this > >> is just the kind of scenario someone evil like Scott would come up with > >> for > >> a question: "R1 and R2 are connected by 100mb ethernet. Pretend it's > >> only > >> 10mb and implement WRD" > >> > >> Thanks, > >> Brian > >> > >> > >> > >> > >> > >> > > > > > > > > > > > -- > Lock the last open door; the ghosts are gaining on me. > > > 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 Oct 04 2008 - 09:26:19 ART