Re: [Ibnemo] some clarification questions about IBnemo

[Zhoutianran <[e-mail]>]

Sue,

You are right. It's interesting to introduce I2RS for the comparison.
I2RS is dynamic compared to YANG.
And IBNEMO sits in a higher level compared to I2RS.



Tianran

From: Susan Hares [mailto:[e-mail]]
Sent: Wednesday, February 11, 2015 9:02 AM
To: Zhoutianran; Georgios Karagiannis
Cc: [e-mail]
Subject: RE: [Ibnemo] some clarification questions about IBnemo

Tianran:

I agree that the static data model for configuration and operational status is the base model from netmod.
However, if one includes the dynamic interface being defined in the I2RS model (pub/sub) then the yang modules could be considered  dynamic supporting temporary configuration and publication/subscription.

All of these are like machine code in the network.  The dynamic nature of I2RS is a descriptive language instead of the prescriptive Intent-based Nemo.   The current 100+ yang modules will have 10,000-20,000 model entities defined.

Nemo's level of dynamic nature is at the intent level which makes it more powerful.  Like a lever, Nemo's intent-based DSL indicate the application wants to move the network. An example of this is a "Big Data" application that wants to control setting up three sites for data transfer in order to do over-night processing for a company,

The 15 primitives of Nemo specify the establishment of a virtual network via (node, link, action, policy)  With the model primitives, you can also create a library of definitions that allow the Big Data Application to pre-set up the names of the nodes or links or flows to be more human readable.  This makes the http:: stream to be human readable which the Network operators prefer.


<model_definition_cmd> := <node_definition> | <link_definition> | <action_definition>


Sue

From: Ibnemo [mailto:[e-mail]] On Behalf Of Zhoutianran
Sent: Tuesday, February 10, 2015 4:53 AM
To: Georgios Karagiannis; Susan Hares
Cc: [e-mail]<mailto:[e-mail]>
Subject: Re: [Ibnemo] some clarification questions about IBnemo

Hi Gergios,

A big difference I noticed is that:
YANG is used for describing static data model including data structure, and operations.
IB-NEMO DSL can be used for programming, i.e. application can interact with the network using NEMO DSL.

There are many other differences I think we need a gap analysis.

Best Regards,

Tianran Zhou (Ph.D.)
Senior Network Researcher
Huawei Technologies Co., Ltd.

From: Ibnemo [mailto:[e-mail]] On Behalf Of Georgios Karagiannis
Sent: Tuesday, February 10, 2015 5:15 PM
To: Susan Hares
Cc: [e-mail]<mailto:[e-mail]>
Subject: Re: [Ibnemo] some clarification questions about IBnemo

Hi Sue,

Thank you very much for the answers. Regarding the listed work items:


    c) Standardize Intent-Based Nemo DSL (domain-specific language) for virtual networks.

        IB-Nemo runs over a RESTful (http) interface.

    d) Standardize Intent-Based Nemo DSL for Security devices

    e) Standardize Intent-Based Nemo DSL for compute devices.




What are the main differences between Intent-Based Nemo domain-specific language and YANG?
By focusing on these three areas, the scope of the WG might be quite large. If it is to focus only on one of these three work items, which one of them will you choose?

Best regards,
Georgios



From: Susan Hares [mailto:[e-mail]]
Sent: Saturday, February 07, 2015 4:15 AM
To: Georgios Karagiannis
Cc: [e-mail]<mailto:[e-mail]>
Subject: RE: [Ibnemo] some clarification questions about IBnemo


Gergios:



Thank you for these great questions.





1) What is the motivation of starting this working group?

The motivation for starting this group comes out three things:  a) desire to maximum usage of Network devices, b) Intent-Based theory and c) an 80/20 rule of interfaces.



Desire to maximum network devices:

Today's networks connect data centers to mobile users.  The Data Centers operate to maximize the use of server's compute cycle, storage devices, and networks.  A research study (2011-2012) examined how to maximize usage of compute, storage, and networks using theoretical models.  This study showed that if you optimize for any one facet (compute, storage, network) you obtain 66% utilization of all facets.  However, if the applications can communicate and negotiate their intent for compute, storage, or network - the theoretical utilization can reach 95% or higher.



The potential higher utilization combined with the lowering of costs due to with the virtualization of network devices creates an economic environment that radically changes the cost dynamics of IT technology.



Intent-Based Networks

The signaling of the network changes via applications that direct SDN-based networks via descriptive commands requires a large data flow between applications and networks.   The application does not really want or need the full description of the networks.   For example, a database application simply needs to signal that it needs to connect to two other database locations and exchange data with some maximum flow rate.



Intent-based network signal this minimum required information, and let the network set-up the rest.



80/20 rules of applications

Some SDN orchestrators are "god-boxes" that need to know and direct everything.  However, most applications really need only to signal intent.   The network engineers designing SDN orchestrators are 20% or less of the total applications.   The other 80% of applications use the 20% of total commands that request the high-level requests.   Intent based Domain Specific languages (E.g. focused on network) allow simple commands to be exchanged between the application and the SDN controller across a RESTful interface (http protocol).



This simpler interface for applications can enable a large class of new applications to utilize the SDN controller.



2) Are there use cases defined that are motivating this work?



The use case motivating this work:



a) Simple interfaces for a new class of monitoring and manage applications for the new NFV controllers controlling Carrier access networks or Cable networks,



b) Simple interfaces for the network-based application store that can load query or load things on mobile networks,



c) Large Database applications desiring to signal the network to set-up a Virtual network to link database sites.  The set-up/tear down of this network is controlled by the Application.



d) Security devices seeking a short-term set-up of application layer connection to exchange data.



3) What are the work items?

     a) Problem statement and use cases document

    b) Intent Framework Description for Networks, Security, Storage, Compute



    c) Standardize Intent-Based Nemo DSL (domain-specific language) for virtual networks.

        IB-Nemo runs over a RESTful (http) interface.

    d) Standardize Intent-Based Nemo DSL for Security devices

    e) Standardize Intent-Based Nemo DSL for compute devices.



4) Is the main goal of the WG to standardize a new protocol, or to extend an existing one?



The main goal is to standardize a series of Domain-Specific languages which run over RESTful interface (http).



5)  What is the relation of IBnemo with existing IETF WGs?



                                       NBI        SDN    SBI

IB-Nemo (application) ---- controller --- I2rs, netconf/netmod,

                                                                       SACM



NBI - north bound interface

SBI - south bound interface



Application protocols work such as  CoApp (CORE) produce http constrained to small environment as a specialized application protocol.  However, these protocols are not focused on enabling an Intent-Based application protocols that will utilize NFV architectures.



Before I launch into more details, please let me know if you have questions.



Sue Hares



-----Original Message-----
From: Ibnemo [mailto:[e-mail]] On Behalf Of Georgios Karagiannis
Sent: Thursday, February 05, 2015 4:32 AM
To: Susan Hares
Cc: [e-mail]<mailto:[e-mail]>
Subject: [Ibnemo] some clarification questions about IBnemo



Hi Sue, Hi all,



In a previous email it is mentioned that:



"The Nemo-project group seeks to create a simple Intent-Based inter-operable application protocol that forms a simple NorthBound API (NB) for applications on any platform to control the following : a) setup and take down of virtual networks between virtual nodes, b) control transfer of data toward storage, and c) handle compute devices with a the minimal set of intent-based primitives."



Can you please provide some clarifications on the following points?



o) What is the motivation of starting this working group? Are there use cases defined that are motivating this work?



o) What are the work items? Is the main goal of the WG to standardize a new protocol, or to extend an existing one?



o)  What is the relation of IBnemo with existing IETF WGs?



Best regards,

Georgios



_______________________________________________

Ibnemo mailing list

[e-mail]<mailto:[e-mail]>

https://www.ietf.org/mailman/listinfo/ibnemo