cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
50
Views
0
Helpful
3
Replies
Cisco Employee

Distinct PE and CPE handling in service model

Hi experts,

I’m brooding over a special customer requirements for NCS integration in brownfield with existing OSS operations.

I fully agree that the environment is sub-optimal, but we should be able to show a way to sneak in there before starting the lengthy cleanup process.

    • The OSS is configuring basic Layer3 PE/CPE services

    • The service description always contains CPE and PE config (e.g. some CPE config relies on data in PE section)

    • The OSS uses elements in the xml data to signal if a deployment or change should be executed on the CPE or PE, never both at the same time

    • OSS expects positive feedback, e.g. for CPE change, before PE change is triggered

I am confident we will survive the POC with an “reconfig CPE and PE” approach, but would need to have an answer, how we would fulfil the above requirements (given no change to OSS).

I currently see the following solution:

    1. OSS sending full service to NCS

    1. Initial service deployment: NCS creating sub-services for PE and CPE (depending on flag in service data)

    1. Service change: NCS updating sub-services for PE and CPE (depending on flag in service data)

Only within the sub-services NED config is created, modified, updated.

Are there other ideas around, excluding re-design of OSS southbound data.

Everyone's tags (4)
1 ACCEPTED SOLUTION

Accepted Solutions
Cisco Employee

Re: Distinct PE and CPE handling in service model

 

Very general question - I'll try though - my 2c

 

 

On 15/07/15 15:20, Mike Leske (mleske) wrote:

 

> Thanks Khara,

 

>

 

> as both topics have been opened by me, let me add a quick note here.

 

>

 

> *YANG data modification by Java code*

 

> Am looking for a general guideline, whether is is good or bad practice

 

> to use the service Java code to set leafs in the service YANG model.

 

 

This is advanced topic, having the service provisioning code manipulate another service. Service A modifies service B.

 

 

Having it change itself is probably no-no - I'm not even sure what that would mean.

 

 

In general - we call this - stacked services. A service which creates or changes another Fastmap service.

 

When the first serv (A) returns, it has changed or created B, we return to the transaction mgr loop inside NSO. It'll see that there are additional new changes, that also have a "service-point"

 

assigned to it - and then it'll run that code too. Until there is nothing left.

 

 

The we continue to calculate the required diffs to /devices/device

 

 

>

 

> *Distinct PE and CPE handling in service model* I receive a private

 

> reply explaining this could be handled within the service Java code.

 

> Here, I¹m also looking for the general guideline, whether the

 

> preferred way would be splitting the complexity into sub services, or

 

> building extensive Java code to handle such special cases.

 

 

 

There is no recommendation here, each case is unique. It's not the case that sub-services == modularity.

 

Modularity resides in the code, not in the HL drawings. It's mostly a matter of how to organize the code.

 

 

In a CPE/PE VPN case, I'd prefer to have one (or a few) templates that cover the entire service, and then let Java code apply those templates. No need to look for stacked services unless absolutely necessary. It just make life more complicated.

 

 

Also - it's not until a second case actually comes up, it makes sense to modularize. I never modularize prematurely.

 

 

/klacke

 

 

/klacke

 

 

 

 

 

View solution in original post

3 REPLIES 3
Cisco Employee

Re: Distinct PE and CPE handling in service model

Hello,

any suggestion, whether sub-services is the right way ? Are there other possibilities known to the community ?

Highlighted
Beginner

Re: Distinct PE and CPE handling in service model

i read two questions: 1) how to position & integrate NCS and the existing OSS? 2) possible and how to make changes to PE & CPE in sequence?

is this understanding correct?

Cisco Employee

Re: Distinct PE and CPE handling in service model

 

Very general question - I'll try though - my 2c

 

 

On 15/07/15 15:20, Mike Leske (mleske) wrote:

 

> Thanks Khara,

 

>

 

> as both topics have been opened by me, let me add a quick note here.

 

>

 

> *YANG data modification by Java code*

 

> Am looking for a general guideline, whether is is good or bad practice

 

> to use the service Java code to set leafs in the service YANG model.

 

 

This is advanced topic, having the service provisioning code manipulate another service. Service A modifies service B.

 

 

Having it change itself is probably no-no - I'm not even sure what that would mean.

 

 

In general - we call this - stacked services. A service which creates or changes another Fastmap service.

 

When the first serv (A) returns, it has changed or created B, we return to the transaction mgr loop inside NSO. It'll see that there are additional new changes, that also have a "service-point"

 

assigned to it - and then it'll run that code too. Until there is nothing left.

 

 

The we continue to calculate the required diffs to /devices/device

 

 

>

 

> *Distinct PE and CPE handling in service model* I receive a private

 

> reply explaining this could be handled within the service Java code.

 

> Here, I¹m also looking for the general guideline, whether the

 

> preferred way would be splitting the complexity into sub services, or

 

> building extensive Java code to handle such special cases.

 

 

 

There is no recommendation here, each case is unique. It's not the case that sub-services == modularity.

 

Modularity resides in the code, not in the HL drawings. It's mostly a matter of how to organize the code.

 

 

In a CPE/PE VPN case, I'd prefer to have one (or a few) templates that cover the entire service, and then let Java code apply those templates. No need to look for stacked services unless absolutely necessary. It just make life more complicated.

 

 

Also - it's not until a second case actually comes up, it makes sense to modularize. I never modularize prematurely.

 

 

/klacke

 

 

/klacke

 

 

 

 

 

View solution in original post

Content for Community-Ad
August's Community Spotlight Awards
This widget could not be displayed.