need help in Applying QOS to an ISP on ASR 1002

[Dr.X]

Hi

i want to ask 

if someone gave me CIR with 900Mbps internet upload & download

the question is being asked is.

when the congestion occurs ( when my bw is full)

where does the congestion occur ??

on my router ?

or ISP router ?

or the internet ?

or other place ???

that issue is deiving me mad , because im doing qos on my side to guarantee bw and shape  but it dont give me a result ?!

i mean , the qos is maching ok and soo nice

but ..... the performance.

agian

the performance as there is no guarantee and no bw guaranteed.

i asked an expert and told me you have to ask  your provider to do that thier side ?!!!  im really shocked to hear that ... he told me that congestion occurs outside my router.

so , im here to ask agian and agian.

what choices i have to fix my issue ??

wish to help

regards

[Reza Sharifi]

Hi,

If you have any monitoring software, you can use it to look at your interfaces connecting to your provider.  If for example you see a constant tx and rx of 800 or 900Mbps then you are pretty much filling up the link and so you need more bandwidth. On the opposite side if you see a constant tx and rx of 100 or 200Mbps, then you know that there is no congestion and you still have room to use more bandwidth.

HTH

[Dr.X]

Hi , thanks I use prtg monitor .

but again I don't want to buy more bw.

i need qos

thsts why I asked that question here

regards

[Evgeniy Prichinin]

So you need implement QoS on ISP side (in place where policing is done) .

[Dr.X]

can you explain why ?

is it impossible do it my side ?

[Evgeniy Prichinin]

For example traffic flow to your network reach 1gbps,  it reach PE router of ISP and it try deliver it to your network but there are policing on interface facing to your equipment which drop exceeded  traffic and you will get only 900mbps. So you need to implement QoS in policing rule on ISP side.

Congestion Caused by Speed Mismatch. PE router have X*10G connections, but your have only 900mbps interface.

[Dr.X]

great , wt about shapping on my side with less than 900 , assume that 890 Mbps has been shaped my side.

i can lose 10 M  so that i apply qos well ,

does that guarantee that congestion is only my side  and let QOS successful on my side ?

agian , im talking about parent , child policy map

cheers

[Evgeniy Prichinin]

Shaping is not solution. Because some traffic already dropped by ISP and speed mismatch still exist. You need implement QoS on ISP side with different rules for different traffic types.
 

[Dr.X]

ok , let me ask u more 2 questions :

Q1

the ISP shapping  on the ISP occurs before or after i exceed my 900Mbps ?

that's why i asked u about shaping my side 890 as an example , and in this case the ISP will not feel congestion and will not drop !

can you expalin it more plz ?

Q2

does that mean , all ISPs in the world must contact their main source provider ans ask it to do QOS on its side ??

i already bought ASR 1002 to support QOS , but that's really a very bad news to hear about QOS  !!!

[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

What you might not understand, traffic is bidirectional, so to manage congestion you also need bidirectional QoS.  Using QoS on your egress is fine, but it only addresses "half" the possible congestion issue; you still need to consider congestion toward you.

Some service provider WAN topologies may provide some QoS features, examples include MPLS and MetroEthernet.  Generally, Internet service providers do not.

As I described in my original post, there are some appliances that can apply some QoS features against inbound traffic (generally they pace/shape return TCP ACKs and/or "spoof" receivers RWIN size to regulate inbound transmissions rates).

"Typical" routers cannot not do similar or cannot do them well.  Regarding not well, you can police low priority ingress traffic to drop it early, or you can shape return ACKs.  I've used both, they work, but not with much precision.

So again, your choices are, convince your ISP, or find another, to configure QoS on their side, obtain more physical bandwidth, or try an appliance designed for this purpose.

[Dr.X]

ok great explanation .

let me ask the opposite question .... why my isp dont think the same thing , and go to the bigger provider and ask him to apply qos ?

if i have 900 Mbps , shoudnt it be CIR 900 and guranteed 100 % to me  ?

now lets think in other idea , now forget me and assume only my ISP provdiver and the provider of my isp  provider.

now my isp provider must go and ask its provider to apply qos ?

and each one ask the other  ????

is that logical ? that each one should ask the bigger ISP to help him ?

i think its impossible to apply qos as you want unless you have marking  from the thing you need to access towards ur end !!

although i got a QOS course with deep details but never someone put the torch on that issues !

again , thanks alot for all clarification  and your time .

regards

[Evgeniy Prichinin]

at big speeds there is no difference between shaping and policing. So ISP use policing, we use config something like that

police cir 10240000 conform-action transmit  exceed-action drop 

Tier 2 / Tier 1 ISP may use qos only in their network. If load exceed normal value it's better to make some changes in routing or add some extra bandwidth to overloaded links. It's not good practice to sell oversubscribed bandwidth, otherwise your customer would change ISP ;-).

[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

At "big speeds", there's a difference between shaping and policing.  However, generally within an ISP, policing is used to restrict bandwidth usage, of a customer, to some agreed upon CIR (when CIR is less than physical capacity).  With policing you don't need to manage any generated congestion (as over rate packets are just discarded).  Policing is also "lightweight" in its resource demands and is often directly supported by the hardware.

I do agree ISPs, that provide SLAs, avoid over subscription because otherwise SLAs might not be meet.  However, if SLAs are not provided, over subscription isn't uncommon.  If too much of the latter is done, customers may indeed change ISPs, but do they have a choice?  Or, how do the different ISP fees compare?  (ISPs do analyze customer impact, i.e. customer retention, with cost of providing the service.  They try to determine what's most profitable, not what the customer would like most.  [Do you have gig to your home yet?  If not, why not?  Is it because it cannot be technically done, or is cost and/or what other ISPs offer in the area a factor? (If you're lucky, maybe you have Google fiber, but even that might be a "loss leader".)])

BTW, at "big speeds", another reason shaping is avoided, and also why over subscription is avoided, is because of the needed queuing.  For example, say you wanted to queue up to 100 ms of traffic.  Well on a 100 Gbps link, that's 10 Gb of data.

[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

ISPs often don't bother with QoS because they would rather sell you more physical bandwidth.  Other WAN cloud providers might support some QoS features because they are often offering to replace "private" p2p link where an Enterprise might already use QoS to manage their bandwidth.  (Again, ISPs, traditionally have not supported any QoS.)

Yes, if your CIR is 900 Mbps, that's what you should get.  But, what's the physical bandwidth of the link?  Is it gig?  If so, often providers will police the traffic and drop overrate traffic.  The advantage of using a CIR, you don't need to make physical link change every time your want increase (or decrease) your "rented" bandwidth.

[rtnet_4820]

Ok, its better for you to define a shaper for up to 900M at your side. so possibly your traffic will not be dropped then at the SP side if it Police your traffic as well.

secondly you define bandwidth allocation for your different traffic classes as per your requirement within that shaper you defined already as a parent one.!!!!