Hierchical Shaping Configuration on Sub-Line Rate Interface (QOS)

[eandrcisco007]

Gentlemen,

We are planning to configure parent policy map that fulfill overall shaping on the interface to force the child policy bandwidth, so that the child policy guarantee to stick to the shaped Bandwidth per circuit i.e. 100Mbps (provided by ISP and its contracted rate) instead of normal speed of the physical interface i.e. 1Gbps, so i wonder if someone could guide me through the configuration as to what needs to be configured/defined.

I wonder if i need to configure the Bc and if yes why...or just leave it as below or more:

Policy-map pqos 100mbps

Class class-default

bandwidth 100000

Appreciated for any input and or guidance in this regard.

[Joseph W. Doherty · Accepted solution]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Is it possible to run your interface at 100 Mbps?  If so, that works much better for QoS bandwidth management that using a shaper.

In any case, a nested policy would be configured something like:

policy-map parent

class class-default

shape average 100000000

service-policy child

policy-map child

class A

..

class B

..

class class-default

..

You would apply the parent policy on the interface.

Yes, you might configure Bc, but knowing what to set it to is a bit advanced.  You might first try the default.

BTW, I suspect shapers don't always account for L2 overhead, so you might need to set the shaper's bandwidth slower, I've found about 15% reduction usually works well.

[eandrcisco007]

Joseph,

Could you please be specific about the advantages of lowering the interface to 100 Mbps and what would be its ramifications if we keep at 1000 MbpS instead?

Also, I wonder if you could provide more details about your statement " I suspect shapers don't always account for L2 overhead, so you might need to set the shaper's bandwidth slower"

I am trying to come up with the final configuration and i would like to consider all factors before final implementation.

Thank you.

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

If your CIR was only 100 Mbps, running your interface at 100 Mbps avoids the need to shape, as the interface and CIR are the same.

Avoiding the need to shape, gets your 100 Mbps to behave like 100 Mbps, as that's actually what the media is running at.  When you use a shaper, it's trying to simulate a slower link by holding back traffic that exceeds a total volume over some time.  Traffic under/before the volume limit is transmitted at full port speed, which means QoS might not prioritize the traffic transmission sequence in the same manner (which may be an issue).

To make the prior issue clear, consider a gig port with a 100 Mbps shaper.  Shapers measure volume over some time period.  To highlight the issue I'm describing, assume the measured time period is 1 minute.  So, if something physically transmitted at full gig, for six seconds, a shaper would hold traffic for the remaining 54 seconds.  This would be especially a problem for time sensitive traffic, such as VoIP.

Of course, in real-world, shapers don't generally use 1 minute time intervals, but many of Cisco's older shapers used to default to 25 ms, now Cisco's newer shapers often default to 4 ms, to mitigate this issue.

The reason why I suspect some (Cisco) shapers don't account for L2, because I noticed it was noted as a special feature on some higher end routers.

[eandrcisco007]

Perfect/ pretty helpfull...Thank you, So in my case, since the CIR is 200Mps,

I wonder if i can configure the parent policy on the interface and the shaper as below :

policy-map HQos-200MBPS

class class-default

shape average 200000000

service-policy CE-EDGE

Shaping Rate=200 Mbps;

Nested CE-EDGE queuing policy forces queuing at sub-line rate

int gig 0/1

bandwidth 200000

ip address xx.xx.xx.xx xx.xx.xx.xx

duplex full

speed 100

ip flow ingress

ip flow egress

load-interval 30

service-policy output HQos-200MBPS

Per provided details and explanations , i change the interface speed to 100Mbps and i wonder if it would make any sense considering the fact that my Shaping Rate=200 Mbps based on above configuration?  What if, I just left it on 1000…instead?

Thank you.

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

If your CIR is 200 Mbps, then you don't want to run the interface at 100 Mbps.

I was only suggesting running at 100 as I thought, from your original post, your CIR was also 100.

[eandrcisco007]

Thank you Joseph,

I was just suspicious,

Would you please check the proposed parent policy and update if it is acceptable?

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

The parent policy looks fine, although as I've already explained, you might want to shape slower if you need to account for L2 overhead.

Also, as previously explained, if your CIR is 200 Mbps, you probably don't want to set speed to 100 Mbps on your Gig interface.  (Further, if possible, I recommend auto for Ethernet interfaces, speed/duplex.)

[eandrcisco007]

Thank you and much appreciate for all your inputs and provided explanation.

I am going to consider 15% reduction on shaper prior final configuration.

Have a wondeful day.

[eandrcisco007]

Can you please provide more details as to why you recommand to change the Ethernet Interface to auto for speed/duplex please specifically in our case?

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

It enables some Ethernet feature that deal with error detection.  It's a required part of gig Ethernet.  It's generally recommended by almost all equipment vendors.

Hard coding speed/duplex is very "old school"/"last century" when auto was brand new, optional with FE, and a bit "iffy", especially between vendors.

[eandrcisco007]

Awesome...thank you...

Is there any link through cisco that explains it in details if you may kindly share?

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Sorry, not off-hand.  However, you can search Cisco's main site, or the Internet generally.

[eandrcisco007]

Thank you for all your valuable feedbacks.

[eandrcisco007]

Thank you Joseph, great stuff...

I have got to correct my statement about bandwidth and it should read as 200mbps per circuit provided by ISP and to answer your question to run the interface at 100 Mbps, the answer is no... it is certainly gona be1gbps on the interface.

According to Cisco books, the bandwidth guarantees are triggered at the shaped bandwidth level, and that is set to be equal to the provider contracted rate instead of the normal speed of the physical interface (1Gbps).

As you have mentioned, this stuff is pretty advanced and i have got a long way to get there so i would rather try it all at the default level.

Thank you.