[Dev] Proposal for modeling QOS
James Brunner
jbrunner at alterpoint.com
Sun Apr 20 02:01:49 CDT 2008
Ok, traffic policing lives within the policymap as a separate bit so how does this look...
<QOS>
<classMap> occurs 0 to infinity
</matchIpDscp>
</matchIpPrecedence>
</matchAccessGroup>
</shapeAverage>
</classMap>
<policyMap> occurs 0 to infinity
</name>
<class> occurs 0 to infinity
</name>
</priorityPercent>
</bandwidthPercent>
</shapeAverage>
</ipPrecedence>
</ipDscp>
</servicePolicy>
<trafficPolice>
</bps>
</normalBurst>
</maxBurst>
</confirmAction>
</exceedAction>
</violateAction>
</trafficPolice>
</class>
</policyMap>
</QOS>
I'd tried to make all my proposed extensions (ntp, mpls etc) vendor-neutral so it was all Common.
Were you only going to extend Cisco?
-- JB.
________________________________________
From: dev-bounces at ziptie.org [dev-bounces at ziptie.org] On Behalf Of Zack Salinas [zsalinas at alterpoint.com]
Sent: 18 April 2008 17:37
To: ZipTie Development List
Subject: Re: [Dev] Proposal for modeling QOS
Awesome, exactly the sort of feedback I was looking for. Excellent point on breaking out the class-maps. The interface level QOS info, I think, would have to be worked in to the Common model stuff instead of Cisco but it would definitely be worth having. Traffic policing is a function of QOS (and it's applicable beyond Frame Relay), but I really don't know enough about that to implement any sort of modeling for it.
Cheers,
-Z
________________________________________
From: dev-bounces at ziptie.org [dev-bounces at ziptie.org] On Behalf Of James Brunner [jbrunner at alterpoint.com]
Sent: Thursday, April 17, 2008 5:41 PM
To: ZipTie Development List
Subject: Re: [Dev] Proposal for modeling QOS
My 1.062p's worth (only valid at today's exchange rate for the next 2 hours)
I was going to make a single addition but I ran one of our client's configs against the proposed model and hit a snag.
Class-maps can be used by zero or more policy-maps. If we break class-maps out of the policy map then it works perfectly.
<QOS>
<classMap> occurs 0 to infinity
</matchIpDscp>
</matchIpPrecedence>
</matchAccessGroup>
</shapeAverage>
</classMap>
<policyMap> occurs 0 to infinity
</name>
<class> occurs 0 to infinity
</name>
</priorityPercent>
</bandwidthPercent>
</shapeAverage>
</ipPrecedence>
</ipDscp>
</servicePolicy>
<trafficPolice>
</bps>
</normalBurst>
</maxBurst>
</confirmAction>
</exceedAction>
</violateAction>
</trafficPolice>
</class>
</policyMap>
</QOS>
Also, you can have one or more policy maps applied to an interface either inbound or outbound.
<Interface>
<QOS>
<PolicyMap> occurs 0 to infinity
</name>
</direction>
</PolicyMap>
</QOS>
</Interface>
Not sure if you can count traffic policing as QOS (does anyone still use Frame/Relay?) but we could add that too...
JB.
-----Original Message-----
From: dev-bounces at ziptie.org [mailto:dev-bounces at ziptie.org] On Behalf Of Zack Salinas
Sent: 17 April 2008 17:42
To: dev at ziptie.org
Subject: [Dev] Proposal for modeling QOS
I think it would be pretty cool to extend the Cisco modeling to support QOS, however QOS isn't my strong point so I wanted to throw out this rough draft of what I think the structure might look like and see if you guys have recommendations for adjustments:
<QOS>
<policyMap>
</name>
<class>
</name>
<classMap>
</matchIpDscp>
</matchIpPrecedence>
</matchAccessGroup>
</shapeAverage>
</classMap>
</priorityPercent>
</bandwidthPercent>
</shapeAverage>
</ipPrecedence>
</ipDscp>
</servicePolicy>
</class>
</policyMap>
</QOS>
-Z
_______________________________________________
Dev mailing list
Dev at ziptie.org
http://mailman.ziptie.org/listinfo/dev
_______________________________________________
Dev mailing list
Dev at ziptie.org
http://mailman.ziptie.org/listinfo/dev
_______________________________________________
Dev mailing list
Dev at ziptie.org
http://mailman.ziptie.org/listinfo/dev
More information about the Dev
mailing list