Re: [Teas] clarification to actn-vn-yang

["Ogaki, Kenichi" <[e-mail]>]

Hi Dhruv,

Thank you for the clarification.

Even though I just looked through the section 7 and doesn't feel intuitive, I understand the situation.

If we require any performance constraint to VN topology, CNC must post te-topo with a connection matrix along with the constraint, in the case of section 3.1.
For example, path-metric-bounds should be set to path-constraints of connectivity-matrices instead of bandwidth-generic in section 7.2 example.
We will check carefully if this meets what we want.

>My initial feeling was we could get it from the bandwidth at the VN-member level but I think it would be still useful to put it in vnap as well. I have added it now.

The -09 version added max-bandwidth under vn-ap entry, but ro (operational) data. What do you intend this? You mean that we should set vn-member level bandwidth constraints with te-topo as well, and just retrieve the ro data from this model..

Thanks,
Kenichi

-----Original Message-----
From: Teas <[e-mail]> On Behalf Of Dhruv Dhody
Sent: Tuesday, July 14, 2020 3:52 AM
To: Ogaki, Kenichi <[e-mail]>
Cc: 丹羽 朝信 <[e-mail]>; [e-mail]; 宮坂 拓也 <[e-mail]>; TEAS WG ([e-mail]) <[e-mail]>
Subject: Re: [Teas] clarification to actn-vn-yang

Hi Kenichi,

Thank you for your comments and insight. 

On Fri, Jul 10, 2020 at 4:34 PM Ogaki, Kenichi <[e-mail] <mailto:[e-mail]> > wrote:


	Hi actn-vn-yang authors,
	
	Can I clarify the following two questions?
	
	
	1) consideration of the key network performance constraints
	
	I'd like to clarify how actn-vn-yang considers key network performance constraints in the sense of RFC8233 Appendix A, latency, delay variation, loss,...?
	
	


We rely on the te-topology connectivity matrix for this. Refer the path-metric-bounds and optimization, there based on the metric-types these can be supported. The common te-types in RFC 8776 include some of them and I would guess other modules would augment more types. 
 

	As an operator, we at KDDI are internally discussing to adopt actn-vn-yang to compute/setup VN (topology) as NBI consumed by our OSS/BSS corresponding to CNC. Then, we ask you to accommodate the consideration of the key network performance constraints into the model, for example at the same level of vn-level-diversity in both a vn-list entity and vn-compute/input.
	
	


This was our approach initially, but the WG did not like that and wanted us to re-use the existing TE topology model, and thus we came up with an approach where a reference to an abstract node (of the TE topology YANG model) is put in VN Yang model and we rely on the connectivity-matrix structure to assign the constraints. See the backup pages in https://datatracker.ietf.org/meeting/105/materials/slides-105-teas-sessb-04-a-yang-data-model-for-vn-operation-00.. I have added some descriptions in the appendix as well in the latest version. 

 

	As the same context, actn-pm-telemetry-autonomics, some authors overwrap, already considers the key network performance data and compute/setup VN as autonomic scaling intent as follows. In that sense, CMI of ACTN must support this capability and it's natural for actn-vn-yang to do so, too.
	
	actn-pm-telemetry-autonomics says:
	2.  Use-Cases
	   o  Customer services have various Service Level Agreement (SLA)
	      requirements, such as service availability, latency, latency
	      jitter, packet loss rate, Bit Error Rate (BER), etc.  The
	      transport network can satisfy service availability and BER
	      requirements by providing different protection and restoration
	      mechanisms.  However, for other performance parameters, there are
	      no such mechanisms.  In order to provide high quality services
	      according to customer SLA, one possible solution is to measure the
	      SLA related performance parameters, and dynamically provision and
	      optimize services based on the performance monitoring results.
	
	3.2.  VN KPI Telemetry Model
	   This module also allows autonomic traffic engineering scaling intent configuration mechanism on the VN
	   level.
	
	4.  Autonomic Scaling Intent Mechanism
	   o  performance-type: performance metric type (e.g., one-way-delay,
	      one-way-delay-min, one-way-delay-max, two-way-delay, two-way-
	      delay-min, two-way-delay-max, utilized bandwidth, etc.)
	
	   o  threshold-value: the threshold value for a certain performance-
	      type that triggers scale-in or scale-out.
	
	


You are correct to note that this model provides the (1) measured telemetry data and (2) specify threshold based on different metric-type on for scaling purposes. This is a new advanced feature defined in this I-D for the first time. 

But, the WG felt that specifying constraints at the time of VN creation can be achieved by the existing connectivity matrix construct and thus we were asked to reuse it.  
  


	2) VNAP level bandwidth control
	
	The current model can indicate the max/available bandwidth to Access Point corresponding to (physical) customer's end point. However, we also want to do so at Virtual Network Access Point level, which corresponds to traffic engineering of vn-member level. Is there any problem?
	
	


My initial feeling was we could get it from the bandwidth at the VN-member level but I think it would be still useful to put it in vnap as well. I have added it now. 

Thanks! 
Dhruv
 


	We of course understand augmentations to solve these requirements, but we believe these are general requirements, and then desire to accommodate to the model.
	
	Thanks in advance,
	Kenichi
	
	-- 
	Kenichi Ogaki
	KDDI Corp. | Operation Automation Promotion Dept.
	[phone] | www.kddi.com <http://www.kddi.com>