WRR QoS unused traffic classes

We are planning QoS for the enterprise LAN and MPLS core. At present, there are 5 traffic classes identified and no VoIP traffic. We could allocate resources and bandwidth for 5 queues and deploy such configs. But I guess it's worth defining all 8 classes and allocate queue limits and bandwidth to them. Later it will be easier to mark the new traffic and classify it into the unused queues without modifying the wrr config.
67xx 1p7q8t line card
Q3 and Q8(PQ) will not be used. 5+15% is allocated to them.
priority-queue queue-limit 15
wrr-queue queue-limit 30 15 5 10 15 5 5
wrr-queue bandwidth percent 30 15 5 10 15 5 5
I'm not sure if the allocated bandwidth and queue limit for the unused queues will affect the existing traffic and limit the aggregated traffic quantity. Will the 5 classes be able to fill the bandwidth until there is no congestion? What happens in case of congestion? Can the traffic excess 80% or not? (For simplicity, 100% is regarded as a fraction of link capacity defined by max-reserved-bandwidth)

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
Generally, unless you're shaping or policing, bandwidth sharing commands provide a minimum bandwidth guarantee, and unused bandwidth can be used by other queues. Also when working with more than one queue, relative ratios are usually preserved. So, for instance, if q1 was configured for 25%, q2 for 25% and q3 for 50%, if q2 had no traffic, and q1 and q3 wanted all they could have, they would split the bandwidth 1:2 or 1/3 to 2/3.

Similar Messages

Setting the traffic class in ServerSocket

Hi all,
in a ServerSocket I want to mark the dscp fields of the packets which are flowing through the client. I set the traffic class field as follows:
Socket inSock = serverSocket.accept();
inSock.setTrafficClass(64);
but this has a problem that the first packet sent by the inSock cannot be marked. This occurs because the first packet is sent in the accept() method, but I set the traffic class after this packet is sent. Then do you know any solution to set the tos as the default (set before the first packet) option for all sockets accepted by a ServerSocket.
I have some very workaround solutions such as implementing the SocketImpl and etc.But I want a KISS (Keep it simple stupid).
Thanks.

Yes thanks for your offer it would be good approach. but this arouses a new problem that my ServerSocket are created by a factory and I need to find a transparent solution in which for all ServerSocketFactory objects my solution works well. Let's see the example.
MyChannel(ServerSocketFactory f) {
ServerSocket serverSock = f.createServerSocket(param1, param2, ....);
Socket inSock = serverSock.accep();
In this case how can I morph the serverSock instance to my MySpecialServerSocket object for example.
Indeed I need only overriding accept method, and I need this to do for an instance not an object.
Do you know any way to do it?
For example we can do a thing as follows:
MyObject ob = new MyObject() {
public overridenMethod() {
};

QOS - VOIP traffic: payload and signalling

Two questions For VOIP traffic,
Q.1. should the payload and signalling be assigned the same COS ?? to avoid losing Signalling traffic.
Q.2. flash upgrades to the phone sets are tftp. Should this traffic be assigned COS=0 or the same cos as the signalling traffic ? Phone flash gets corrupted if some of these packets are lost ?
thanks for your help,

Hello,
best practice puts signaling into another class than voip "payload" as you named it.
The reason is that the qos requirements are different. Signaling needs guaranteed bandwidth, voip needs low delay and guaranteed bandwidth.
TFTP can be placed into signaling class or in a separate class.
An example config could look like this:
ip cef
class-map match-any VoIP
match ip dscp ef cs5
match protocol rtp audio
class-map match-any VoIPsignal
match ip dscp cs3
class-map match-all TFTP
match protocol tftp
policy-map VoIPprio
class VoIP
priority percent 10
class VoIPsignal
bandwidth remaining percent 5
class TFTP
bandwidth remaining percent 5
class class-default
fair-queue
random-detect
interface Serial0
ip address ...
service-policy output VoIPprio
You would have to adjust the bandwith ratios for your needs and the interface naming to your environment.
The policy will give 10% of the link to voip traffic with lowest possible delay, 5% of the rest for voip signaling, 5% to TFTP and the rest for the other data present.
Hope this helps

How to apply LLQ QoS to traffic generate by router?

There is a voice gateway at the remote site, the voice card connect to the local pstn, for qos, the router must prioritize the voice traffic which is generated locally.
So is there any idea?

No, it doesn't work. please see below:
R2#sh ip local policy
Local policy routing is enabled, using route map QoS
route-map QoS, permit, sequence 10
Match clauses:
ip address (access-lists): IP_QOS_5
Set clauses:
Policy routing matches: 11060 packets, 950804 bytes
Local Policy match many packet of IP prcendence 5 or RTP traffic. But... CBWFQ only match a little RTP packet...
Class-map: Voice (match-any)
5 packets, 458 bytes
30 second offered rate 0 bps, drop rate 0 bps
Match: protocol rtp
5 packets, 458 bytes
30 second rate 0 bps
Queueing
Strict Priority
Output Queue: Conversation 136
Bandwidth 70 (%)
Bandwidth 358 (kbps) Burst 8950 (Bytes)
(pkts matched/bytes matched) 0/0
(total drops/bytes drops) 0/0

ASA5505 - Qos / Priority / Traffic Shaping - VOIP/SIP

I have a client using a VOIP service to a third party provider (RingCentral). They are connected via Cable ISP (6mb) to the Internet and now experiencing performance issues with their VOIP service. They indicated that the call can be heard but that there is jitter and choppines in the call and they have to place a regular landline call. Their provider recoomended using QOS to help improve. I did not see anything straight forward on the ASDM interface to do this and figure it may require command line to accomplish.
They have Cisco IP 303 and 5252G2 phones which connect through an ASA5505 7.2(4) to their provider for service. Apparently the voip app uses the following ports:
UDP
5060-5090
8000-8200
16384-16482
What would be the best solution to improve performance or perhaps traffic shape / priortize traffic to help. I assume this may be happening if there are heavy downloads or activity happening on the network. The ASA5505 is on 7.2(4). I'd appreciate if someone could provide some coded examples for the above info.
Much appreciated!

Hi Vito,
You can prioritize your voice traffic over data traffic, refer to this doc for prioritization:
http://www.cisco.com/en/US/products/hw/vpndevc/ps2030/products_tech_note09186a008084de0c.shtml#Pqu
Hope that helps
Thanks,
Varun

QoS Marking Traffic Generated by Network Device

I am working on defining QoS configuration standards for Catalyst 3750, 4500 and 6500 platforms. I would like to provide preferred treatment to some of the protocols commonly used for network management (i.e. telnet, ssh, etc...). I have no issues with classifying and marking traffic generated by a NMS or PC connected to a switch but I can not figure out how to classify and mark traffic generated by the switch itself. It won't do any good if traffic from the NMS or PC to the device is preferred but the response from the device is unmarked/best effort.
The only way that I can see to do this is to apply an input policy on the upstream device to classify and mark the network management traffic. I would prefer to do it on the device itself. Does anyone know if that is possible?
I'm assuming that any traffic generated by the switch would be ignored by an output policy, similar to the way that traffic generated by the device will by-pass access-lists. Even if this is a bad assumption the 3750 does not support output policies, so I'd still be stuck there.
It was suggested that I look into applying the policy to the control plane however on the platforms that options is available, policies can only be applied to the input. Which I believe is the wrong direction.
Anyone have other suggestions on how to accomplish this or is this a lost cause?

You can use local PBR to do this. Check out the local policy-map
http://www.cisco.com/univercd/cc/td/doc/product/software/ios122/122cgcr/fqos_c/fqcprt1/qcfpbr.htm#wp1001002

How to priorities(QoS) the traffic for DSCP 46 and 34 in cisco 2960s switch

HI,
We are going to implement Microsoft Lyncs 2013 in our network, so how to priorities the traffic for DSCP 46 and 34 in cisco 2960s switch. Kindly replay with detailed QoS commands for enabling QOs in LAN.
Thanks
Sujish

Hi,
Have a look at this config guide for all the details:
http://www.cisco.com/c/en/us/td/docs/switches/lan/catalyst2960/software/release/12-2_53_se/configuration/guide/2960scg/swqos.html
HTH

QoS - realtime traffic must be less than one-third of link bandwidth capacity

http://www.cisco.com/c/en/us/td/docs/solutions/Enterprise/WAN_and_MAN/QoS_SRND_40/QoSWAN_40.html
Below is an extract from this document:
Cisco Technical Marketing has done extensive testing and has found that a significant decrease in non-realtime application response times occurs when realtime traffic exceeds one-third of link bandwidth capacity. Extensive testing and customer deployments have shown that a general best queuing practice is to limit the amount of strict priority queuing to 33% of link bandwidth capacity.
Have people found this to be their experience?
IE. have you seen cases where realtime traffic is > 33% of link bandwidth and the non-realtime traffic gets badly affected?
Thankyou, MH

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
"misunderstanding" - yes and no
No, you're correct about Cisco's warning. If you have a 3 Mbps, and assigned 2 Mbps to LLQ and it's using it, effectively you have a 1 Mbps link for your remaining traffic. This is what Cisco is warning about. But, whether that would be a real issue or not, depends on the needs of that other traffic now reduced to only 1 Mbps.
Yes, you might misunderstand about "3x". This is something I'm warning you about, not Cisco. If your LLQ traffic "needed" 2 Mbps, but there's only 3 Mbps total, some LLQ traffic may suffer. It too depends on the actual needs of your LLQ traffic. For something like CBR VoIP g.711, this may not be a problem, but for VBR traffic, like VoIP g.729, or many video codecs, this can be a problem. Basically, you need bandwidth to borrow from during transient bursts. (Also, LLQ normally has an implicit policer, but policers also allow bursts, so again, you need to insure there's sufficient "excess" bandwidth.)
The worst case example would be 2 Mbps LLQ on a 2 Mbps link. I'm sure you can understand "other" traffic might be starved of all its bandwidth, but the LLQ traffic, if VBR, will often behave badly too.

Does Huawei router NE40 support Class-Based QoS?

As I know Class-based QoS defines traffic classifiers based on certain rules and associates traffic classifiers with certain traffic behaviors, forming certain traffic policies. After
these policies are applied to interfaces, class-based traffic policing, traffic shaping, congestion management, and precedence re-marking are implemented.
Does Huawei router NE40 support Class-Based QoS？

The NE80E/40E supports DiffServ and provides standard forwarding services such as EF and AF for users by using the following traffic management measures:
1 Traffic classification
2 Traffic policing
3 Traffic shaping
4 Congestion avoidance
QoS of the NE80E/40E supports traffic policy with the above measures and mapping between the QoS fields in the IP header and the MPLS header.
And more information about router NE40, please visit:
http://www.huanetwork.com/huawei-router-ne40e-series-price_c89

Bandwidth allocation | default class|CBWFQ

Hi everybody
Let say we have 100 mig circuit. Max -reserved bandwidth is 100 mig as well. We make following allocations:
Class A
bandwidth 20
Class B
bandwidth 60
Class Default.
1)We did not make any bandwidth allocations for default class. Assuming We are congested ( i.e class A, class B ), What is the maximum bandwidth Class Default can use?
2) Let say we are congested ( classA,classB) but there is no traffic in default class. How will this unused 20 mig will be distributed among these classA and class B?
++++++++++++++++++++++++++++++++++
I am getting confusing answers:
For example:
From one of theblog ( Dont want pick on author so did not quote it)
You'll want to configure a bandwidth command under the class class-default Otherwise, IOS will divide any unallocated bandwidth equally among all classes; this can result in the class-defaulthaving a very small amount of bandwidth.
Cisco QOS Documentation says:
http://www.cisco.com/en/US/docs/ios/qos/command/reference/qos_cr.pdf
From the above link:
The following output from theshow policy-map interfacecommand on serial interface 3/2 shows that 500 kbps of bandwidth is guaranteed for the class
named voice1. The classes named class1 and class2 receive 50 percent and 25 percent of the remaining bandwidth, respectively. Any unallocated bandwidth
is divided proportionally among class1, class2, and any best-effort traffic classes
Which One is true statement?
If Cisco documentation is correct, then what proportion of unallocated bandwidth is given to default class as there is no bandwidth percentage configured under default class.
Thanks

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
#2 If there's just traffic for classes A and B, they will proportionally share it 20:60 or 1:3 (or 25%:75%). (NB: the former is assuming they want "more". If not, actual usage might not reflect bandwidth allocation. For example, if class B only used/wanted 50% of link, class A could obtain the other 50%.)
#1 recall finding some blog that really went deep into what class-default gets when you don't explicitly allocate bandwidth. I also recall it may have been IOS version dependent and whether FIFO or FQ was defined in class-default (this also was pre-HQF).
It was very complicated, and IMO, best avoided by defining bandwidth in class-default, if class-default usage, relative to other defined classes, is important to your QoS policy.
Generally, if something isn't clearly documented as expected behavior, I avoid relying on "discovered" actual behavior, because it might change with next IOS release.

QoS removed from line :@ Getting to the point of c...

Hello,
After a week of getting BT to fix my BT Vision service which they did and enabled QoS.
I am ****** off to say that the QoS has been removed from our line and again we cant watch anything at all On Demand without the dreaded VO4 error!!
It's getting to the point of us jacking in the TV package and go to Sky for TV services unless it's fixed within the next 48 hours...
FAQ
<script type="text/javascript">// paintProgressAndMessageOnBar(100,"The test has successfully completed"); // </script> Test1 comprises of three tests
1. Best Effort Test: -provides background information.
Download Speed
9454 Kbps
0 Kbps
21000 Kbps
Max Achievable Speed
Download speedachieved during the test was - 9454 Kbps
For your connection, the acceptable range of speedsis 4000-21000 Kbps.
Additional Information:
Your DSL Connection Rate :11192 Kbps(DOWN-STREAM), 1107 Kbps(UP-STREAM)
IP Profile for your line is - 9000 Kbps
The throughput of Best Efforts (BE) classes achieved during the test is - 15.68:22.66:61.67 (SBE:NBEBE)
These figures represent the ratio while sententiously passing Sub BE, Normal BE and Priority BE marked traffic.
The results of this test will vary depending on the way your ISP has decided to use these traffic classes.
2. Upstream Test: -provides background information.
Upload Speed
827 Kbps
0 Kbps
1107 Kbps
Max Achievable Speed
>Upload speed achieved during the test was - 827 Kbps
Additional Information:
Upstream Rate IP profile on your line is - 1107 Kbps
Thanks

There is an issue with BBC iPlayer software update which has buggered up the box
It isnt very stable at all, it's all well and good having iPlayer but if it doesnt work properly I want the new version fixed or the old software back and lose access to BBC iPlayer.
Nothing wrong with the QoS at all- I have had to factory rebuild the software 5 times since the iPlayer update on Thursday morning, box keeps locking up and if you rewind/ ffw anything on demand you get VO4- wasnt an issue before the iPlayer release.
The box is about the same speed as the antiquated Scientific Atlantic V+ HD Box which is slow...

Class-Map and Policy-Map Configuration in CM Confusion

Hi,
I'm implementing a green field WAAS deployment for a customer. We currently have a Proof-of-Concept up and running.
I've got some questions regarding custom class-map and policy-map configuration in the CM. I'd like to nail-down the custom class-map and policy-map configuration (and understanding) in the PoC before cutting over the PoC branches to the production WAAS environment.
Assuming a typical WAAS Deployment using WCCP for off-path interception, branch to DC.
==> 61 in LAN (BRANCH ROUTER) <== 62 in WAN (WAN CLOUD) ==> 61 in WAN (DC ROUTER) <== 62 in LAN
We are using two distinct device groups, BRANCH and DATA CENTER.
If the customer has traffic that we need to classify in order to provide TFO only optimisation, should the single class-map include the traffic in both directions? Ie., (assume the SERVER is 10.1.1.1 TCP Port 443). Should the class-map be configured as:
Class-Map
Line 1: DST IP 10.1.1.1 DST Port 443
Line 2: SRC IP 10.1.1.1 SRC Port 443
Or in this case is only the DST line required? And in which Device Group should the custom policy be applied? Or should it be applied to both Device Groups? If it should be applied to both Device Groups, then would it make more sense to have the policy-map in the Branch DG configured to match the DST traffic, and on the Data Center DG have a different class-map match the SRC traffic?
My confusion is how to classify the traffic (SRC or DST or Both - Separate classes for each or different lines within the same class-map), and where to apply the appropriate policy (both Device Groups, just Branch, just DC) and why...
I tried to apply a custom policy and the impact in the PoC was that the TCP Summary report stopped reporting the individual traffic classes showed 'other traffic' only. Can anyone explain why this may have occurred?
I hope this makes sense.

for instance like this:
policy-map police-in
class class-default
police rate 10 mpbs <optionally set burst>
policy-map shape-out-parent
class class-default
shape 10 mpbs <optional burst config>
service-policy shape-out-child
policy-map shape-out-child
class class-default
queue-limit 10 packets
int g 0/0/0/0
service-policy police-in in
service-policy shape-out-parent out
also have a look at CL 2013/2014 (orlando/sanfran) ID 2904 for more QOS details
and the support forum article of "asr9000 quality of service architecture"
xander

Priority queue for voice/audio traffic

Hi,
Still in limbo after multiple discussions with our vendors, TAC and in general other engineers, so starting a thread here. In the process of rolling out enterprise audio, with the intent to prioritize and allocate 25% of link bandwidth for voice class.
Our config snapshow is as follows -
policy-map qos-wan-out
class dscp-voice-lan
set ip precedence 5
priority percent 25
I understand that
-DURING congestion, this will ensure voice gets a maximum of 25% and is dequeued first due to the priority setting
-And during NO congestion, the voice traffic will be dequeued before other traffic, but at the same time, can go over 25% as QoS kicks in only during congestion.
I am seeing some contradictory results in that we are having high packet loss if we exceed 25% even when the link is less than 40% utilized. I doubt the above CE configurations are an issue. But, wanted to run this by this group.
Alternate theory is that with the above configurations, our traffic is exiting fine - but the service provider who is using priority class queuing within their MPLS network may be capping the bandwidth at 25% at all times (with or without congestion).
thanks

Hi Bro
Maybe the incoming voice packets into your FW isn't marked with ef. For this reason, you don't see anything at all. I hope the QOS isn't tied to a subinterface, as QOS is only supported on the main interface itself. What you're doing here is QoS Configuration based on DSCP. You could refer to this URL for troubleshooting purposes.
http://www.cisco.com/en/US/products/ps6120/products_configuration_example09186a008080dfa7.shtml#tab4
Did you marked on the Cisco Catalyst switchports, which ports are ef?

DCBX Between Linux and Nexus QOS

Hi Guys
i have a server with 2 physical links only.
over this link im passing 2 vlans in trunk ( can be Portchannel to VPC or bond active passive , doesnt matter to me ) .
vlan 302 is for data
vlan 303 is for backup
server on vlan 303 is pushing traffic on that link and would take as much
bandwidth as we would give it , and would therefore cause communication issues to vlan 302 . ( congestion ) .
I would like to talk about my options here , to avoid congestions .
thought about enabling dcbx on clients ( linux ) and pfc ( priority flow
control ) to be able to negotiate qos and bw allocation according to COS IEEE 802.1p .
the idea is to have either the traffic prioritized/limit/shaped before/when its ingressing the nexus port .
the dcbx looks very cool like i said , it's kind of like what the nexus has with the CNA Card .
when it can send the linux nic through dcbx the qos parameters to apply ! ( ah ok , i see now its exactly the same )
ps linux dcbx link : http://manpages.ubuntu.com/manpages/precise/man8/dcbtool.8.html
Ingress Queuing Policies
You can associate an ingress policy map with an Ethernet interface to guarantee bandwidth for the specified traffic class or to specify a priority queue.
The ingress policy is applied in the adapter to all outgoing traffic that matches the specified CoS value.
When you configure an ingress policy for an interface, the switch sends the configuration data to the adapter. If the adapter does not support the DCBX protocol or the ingress policy type-length-value (TLV), the ingress policy configuration is ignored.
did anyone here tried it or implemented dcbx before ?
anyway id be happy to hear your wise and helpfull suggestions please
Thanks Thanks !!!

Found the answer.
Anatomy of RX packet counters on a Cisco Nexus 5548:
Nexus5548# show int eth1/29
RX
6387682660 unicast packets 1495485 multicast packets 164 broadcast packets
6389178309 input packets 589693485138 bytes
5146969 jumbo packets 0 storm suppression bytes
* "input" = unicast + multicast + broadcast (all non-jumbo)
* "jumbo" = frames sized 1519 bytes to the MTU bytes defined on the interface
* "jumbo" frames are counted separately from non-jumbo ("input")
* (unicast, multicast and broadcast "jumbo" frames appear to be rolled into the
one "jumbo" counter)
This can be seen another way with the 'count detail' command, correlated to "show int" output:
Nexus5548# show int eth1/29 count detailed
Ethernet1/29
Rx Packets: 6389178309 <-- input
Rx Unicast Packets: 6387682660 <-- unicast
Rx Multicast Packets: 1495485 <-- multicast
Rx Broadcast Packets: 164 <-- broadcast
Rx Jumbo Packets: 5146969 <-- jumbo
Rx Bytes: 589693485138
Rx Packets from 0 to 64 bytes: 1059418 <-- non-jumbo start
Rx Packets from 65 to 127 bytes: 6072007478 +
Rx Packets from 128 to 255 bytes: 11055520 +
Rx Packets from 256 to 511 bytes: 287107435 +
Rx Packets from 512 to 1023 bytes: 12801466 +
Rx Packets from 1024 to 1518 bytes: 23 <-- non-jumbo end
Rx Trunk Packets: 6387682823 <-- unicast + broadcast
fyi.

SNA and QOS

Subject: SNA/QOS
I AM CURRENTLY RUNNING 12.1(11a) IOS ON A ROUTER BASED NETWORK WITH FOUR
7500 ROUTERS. I AM CURRENTLY RUNNING CUSTOMER QUEUEING AS MY QOS. I AM
NOW LOOKING TO USING LLQ W/ CBWFQ AND DISTRIBUTED MODES, AS WELL. I HAVE
VIP-40 CARDS IN ALL ROUTER. HOWEVER, MY NETWORK IS COMPOSED OF --
TP0 - BATCH TRAFFIC
TP1- INTERACTIVE TRAFFIC
TP2-CONTROLLED TRAFFIC - HIGH AMOUNTS
I HAVE BEEN TOLD THAT CISCO'S QOS WILL SUPPORT TP1 TRAFFIC SO, I AM
CONCERNED ABOUT MY TP2 (CONTROLLED TRAFFIC) AND IF I CAN IMPLEMENT LLQ/CBWFQ
INTO MY NETWORK.
I WOULD LIKE TO IMPLEMENT THIS FOR MY CRITICAL APPS SUCH AS SNA, AND HTTP
WEB BASED APPS. THUS, LIMITING MY SQL, FTP AND TELNET TRAFFIC.
THANKS IN ADVANCE.
ANY PPT FILES WOULD BE APPRECIATED.

Hi Connie,
You open the door to an interesting discussion, by including both SNA and QoS in your question. As you know, a variety of QoS mechanisms have been available for SNA traffic for quite a while. I can't tell from your description what sort of traffic is contained in "TP2-CONTROLLED", so I will assume it is voice and video traffic, and not the "TP2" from OSI Transport Protocols.
The quick answer is that you can definitely support TP1 traffic while protecting the quality of your TP2 traffic. Essentially you decide what percentage of the available bandwidth to allocate to each class of traffic, providing a minimum guaranteed value. In addition, there is a special class, Low Latency Queuing (LLQ), available for delay and jitter sensitive traffic such as voice. Within each traffic class, IOS will then provide Weighted Fair Queuing (WFQ) for each unique flow (session). This becomes a bit more interesting when combined with different WAN types such as frame relay and ATM, you can take advantage of the VIPs that you mention, and there are considerations for low speed circuits. So here are a couple of URLs that provide more information.
http://www.cisco.com/univercd/cc/td/doc/product/software/ios121/121cgcr/qos_c/qcprt2/qcdconmg.htm
http://www.cisco.com/univercd/cc/td/doc/product/voice/ip_tele/avvidqos/qoswan.htm
Coming back to the SNA traffic, I assume you're currently using custom queuing with DLSw+, using the priority parameter to create four TCP connections, and classifying traffic using one of the three available methods. In moving to a CBWFQ model, you will want to map the existing custom queues into the newly created classes. In other words, you can continue to use the same classification techniques, while changing to the easier to define, and more efficient WFQ for output queue processing.
The absolute best traffic classification comes with using the Enterprise Extender (EE) feature of SNASw. When SNA traffic is sent across an EE link, the precedence bits in the IP packets are automatically marked with the same values that are used in the SNA Class of Service (CoS). Since SNASw is our APPN node implementation, propagating the precedence markings from SNASw to DLSw+ also provides an automatic means of classifying the SNA traffic.
Rgds, Dan

WRR QoS unused traffic classes

Similar Messages

Maybe you are looking for