答复: Tunnel Design Philosophy

Lizhenbin <lizhenbin@huawei.com> Mon, 02 November 2015 07:16 UTC

From: Lizhenbin <lizhenbin@huawei.com>
To: Aijun Wang <wangaijun@tsinghua.org.cn>, "rtgwg@ietf.org" <rtgwg@ietf.org>, "netmod@ietf.org" <netmod@ietf.org>
Subject: 答复: Tunnel Design Philosophy
Thread-Topic: Tunnel Design Philosophy
Thread-Index: AdEFapt5Fqf1Bn2CSemDcxAR3HYI/AAEY6RwA/ATJZE=
Date: Mon, 02 Nov 2015 07:16:43 +0000
Message-ID: <5A5B4DE12C0DAC44AF501CD9A2B01A8D8CA5FF9F@nkgeml506-mbx.china.huawei.com>
References: <004a01d1057c$47e02310$d7a06930$@org.cn>
In-Reply-To: <004a01d1057c$47e02310$d7a06930$@org.cn>
Accept-Language: zh-CN, en-US
Content-Language: zh-CN
Content-Type: multipart/alternative; boundary="_000_5A5B4DE12C0DAC44AF501CD9A2B01A8D8CA5FF9Fnkgeml506mbxchi_"
MIME-Version: 1.0
Archived-At: <http://mailarchive.ietf.org/arch/msg/rtgwg/Q4D8T5XaRuqqFz80NG5XhOjCskQ>
Precedence: list

Hi Aijun,

I think your tunnel philosophy is reasonable. But there may be challenges in the real implemention to support the philosophy. The challenges are as follows:

1. There are too many types of IP tunnels such as IPv6/IPv4 over IPv4 tunnel, GRE Tunnel, IPSec/IKE Tunnel, L2TP Tunnel,etc. And now NVO3 work proposes

more IP tunnel types such as VXLAN Tunnel, NVGRE, GPE, MPLS in UDP, etc. Though these IP tunnels may share common aspects, they may have essentially

different usages which is does matter.

2. Different IP tunnels may need more pre-configuration and operational data which are different from each other which is difficult to be accommodated in the

module. But when configure these tunnels, these pre-configuration has to be provided firstly. So the tunnel common modules has to intereract with other tunnel

implementation modules.

3. Common Tunnel modules may need more interaction with modules implementing different types of modules. The complexity may increase as the number of

tunnel types. It may need very smart people to understand all possible types of IP tunnels for implementation of the tunnel modules.
4. All these IP tunnel types do not emerge all at once. Even the possible common attributes shown in your your models did not emerge at the beginning. So it is

very difficult to change the existing implementation to abstact the possible tunnel common modules.

In fact we have ever tried your method at the beginning and at last we gave up since nobody could take the challenging work.

Best Regards,

Zhenbin(Robin)

________________________________
·¢¼þÈË: rtgwg [rtgwg-bounces@ietf.org] ´ú±í Aijun Wang [wangaijun@tsinghua.org.cn]
·¢ËÍÊ±¼ä: 2015Äê10ÔÂ13ÈÕ 13:58
ÊÕ¼þÈË: rtgwg@ietf.org; netmod@ietf.org
Ö÷Ìâ: Tunnel Design Philosophy

Hi, RTGWGer and NETMODer:

Here I want to ask for advices from any expert that is familiar with the usages and designs of various tunnel technologies that are wide deployed within the network.
What is the principle and philosophy about the design of Yang Model for these tunnel technologies?

Currently, there are several drafts that has touches this area, but there are some confusions about their designs, for example:
1. Can we organize these tunnel related-Yang models under one common tree?
2. What is the relationship between the tunnel related-Yang model and the interface Yang Model?

Our opinion is that Yang Model is one design tool/language used to standard the interface between the service provider and device(Device Yang Model), and between the service provider and their customer(Service Yang Model), then the design of them should from top to down, find the general aspects of every model branch first and augment them with specific technology later. This seems more common to all the Model/Object design language.

So, for above two questions, we recommend to design one general tunnel-related Yang model that augments from the interface Yang model, and expand to it to cover the various specific tunnel technologies. Doing so has the following benefits:
1. we can focus first the common characteristic of tunnel technology, especially the static tunnel technologies(dynamic tunnel for example MPLS-TE tunnel is the exception)
2. the appearance of the tunnel on router/switch are all one kind of interface. If it augments from the interface tunnel, it can inherit many variables of the interface Yang model.(several drafts have shown their overlapping design of these variables.)

So, how about your opinion and the reason to do them?
Wish can hear more valuable suggestions on the design of the Tunnel-related Yang Model.

Current available drafts about the Tunnel ¨Crelated Yang Model are bellows:
1. https://tools.ietf.org/html/draft-ietf-l2tpext-keyed-v6-tunnel-yang-00
2. http://datatracker.ietf.org/doc/draft-wwz-netmod-yang-tunnel-cfg/
3. http://datatracker.ietf.org/doc/draft-wilton-netmod-intf-vlan-yang/
4. http://datatracker.ietf.org/doc/draft-liu-rtgwg-ipipv4-tunnel-yang/
5. http://datatracker.ietf.org/doc/draft-li-rtgwg-utunnel-yang/
6. https://tools.ietf.org/id/draft-ietf-teas-yang-te-00.txt( This draft is one exception, and seems can¡¯t be generalized with other five drafts)

Best Regards.

Aijun Wang

China Telecom Corporation Limited Beijing Research Institute
Intelligent Network Product Line