Dear Hideki,
Please see comments inline.
On 28.10.09 03:41 "hideki.okita.pf at hitachi.com"
<hideki.okita.pf at hitachi.com> wrote:
> David, Dan, Juergen,
>
>
> Thank you so much for your reviews and comments.
>
> First, please let me summarize David's concerns.
> Please tell me if I have any misunderstanding.
>
> 1. The proposed model seems similar to the information model
> implicitly defined in the existing ENTITY-MIB.
> 2. The draft does not mention the relationship between the proposed
> model and the ENTITY-MIB.
> 3. The draft lacks the concrete definition how the NMS presents
> the virtual network information to NMS clients.
>
> For the first concern, I examined the ENTITY-MIB since I have not
> considered it. I checked RFC 4133 and implementation by a router.
>
> Then, I found that our model have a similarity with the ENTITY-MIB.
> Both of following objects describe the structure of a physical entity.
> - entPhysicalTable and entPhysicalContainsTable in the ENTITY-MIB
> - PhysicalNode, PhysicalNodeGroup, and PhysicalInterface in our model.
>
> However, IMO, the model of the ENTITY-MIB is insufficient to
> manage virtual networks in data centers from following reasons.
> a. ENTITY-MIB does not have connection objects for describing the
> connections between physical nodes.
Yes, that is not the purpose of the ENTITY MIB.
However, topology detection of an IP network is a very common task
of network management systems and well supported by other MIB modules.
I don't think we need new managed objects for this purpose.
> b. ENTITY-MIB does not have connection objects for describing the
> connections between virtual nodes, for example, the connection
> between a virtual switch and virtual machines on a physical server.
This may not be necessary, depending on how you build your information
model. What I think is really missing is a mapping of virtual entities
to physical entities. The virtual entities need relations to their
virtual interfaces and virtual interfaces need to be mapped to real
interfaces.
> c. ENTITY-MIB does not have virtual-virtual relationship objects for
> describing the structure of a virtual entity.
Yes, but again, there may be sufficient other means than the ENTITY MIB
to provide this information.
> Therefore, I'd like to propose our model as an solution and opsawg's
> new work.
>
> For the second concern, I'd like to add a "relationship to the
> ENTITY-MIB" section that describes above distinction to our draft.
This is probably the best way to achieve your objectives.
As said before, I would consider the most important step to find
a good way of mapping virtual entities to real entities.
Best regards,
Juergen
> For the third concern, I can add a section about a MIB module or
> an XML schema to our draft to describe the concrete datamodel.
>
>
> Best regards,
>
> Hideki Okita
>
>
>> I also had a look at the draft and I think that the basic idea
>> of mapping virtual to real resources is very good.
>>
>> Can't we easily align the draft with the ENTITY MIB data model?
>>
>> I could imagine a MIB module based on the ENTITY MIB that realizes
>> the information model for virtual entities in SMI.
>>
>> Juergen
>>
>>
>> On 26.10.09 13:51 "David Harrington" <ietfdbh at comcast.net> wrote:
>>
>>> Hi,
>>>
>>> I did a quick read of this document. I have a concern.
>>>
>>> The IETF decided years ago that a MIB module was the preferred
>>> mechanism for explicitly documenting an IETF standards-track data
>>> model and (implicitly) the IETF standards-track information model.
>>>
>>> Once the IETF began to accept that having only one data modeling
>>> language (the SMI) might not be adequate, the IETF struggled with the
>>> question of "how do we deal with multiple modeling languages?", and we
>>> started to pay attention to the distinction between data model and
>>> information model. RFC3444 was published as output of the NMRG to help
>>> clarify the distinction.
>>>
>>> The informational model in
>>> http://www.ietf.org/id/draft-okita-ops-vnetmodel-01.txt is similar to
>>> the information model that is implicit in the ENTITY-MIB data model
>>> design.
>>>
>>> I would not like to see anybody make the mistake of thinking that this
>>> Informational document represents IETF consensus on an information
>>> model for managing virtual networks. This document needs to be very
>>> clear that the IETF already has consensus on an (implicit) information
>>> model for managing physical entities, logicial entities, and the
>>> mappings between them.
>>>
>>> I think defining an information model for the management of virtual
>>> networks is very important work, and I think this document really
>>> misses an opportunity to help **standardize** how virtual network
>>> management systems present information to NMS clients. This document
>>> would be terrific if it extracted and documented the information model
>>> implicit in the existing IETF standard ENTITY-MIB, and went on to
>>> discuss whether the existing information model does or does not meet
>>> the needs of real-world implementations/deployments.
>>>
>>> If the existing implicit information model does meet real-world needs,
>>> then I think such a document might become a BCP or an Informational
>>> guideline for managing virtual networks. If the existing implicit
>>> information does not meet real-world needs, then the document could be
>>> a justification for an IETF standards track information model that
>>> meets needs better than the current implicit information model.
>>>
>>> Unfortunately, the current document does not even mention that there
>>> is an information model implicit in the ENTITY-MIB standard, which
>>> makes me question whether this document should be published as an RFC
>>> without a clear statement that it ignores, and may be in conflict
>>> with, existing IETF standards.
>>>
>>> dbh
>>>
>>>
>>>> -----Original Message-----
>>>> From: opsawg-bounces at ietf.org
>>>> [mailto:opsawg-bounces at ietf.org] On Behalf Of
>>>> hideki.okita.pf at hitachi.com
>>>> Sent: Monday, October 26, 2009 7:33 AM
>>>> To: sob at harvard.edu
>>>> Cc: opsawg at ietf.org
>>>> Subject: Re: [OPSAWG] call for opsawg agenda time requests
>>>>
>>>> Scott,
>>>>
>>>>
>>>> I'd like to have a slot at the opsawg session in Hiroshima
>>>> to take a presentation about the update of our network
>>>> virtulization model.
>>>>
>>>> "Virtual Network Management Information Model"
>>>> http://www.ietf.org/id/draft-okita-ops-vnetmodel-01.txt
>>>>
>>>> Regards,
>>>>
>>>> Hideki Okita
>>>>
>>>>
>>>>> please send requests for agenda time slots for the opsawq
>>>> session in Hiroshima
>>>>> to the opsawg list asap
>>>>>
>>>>> Scott
> _______________________________________________
> OPSAWG mailing list
> OPSAWG at ietf.org
> https://www.ietf.org/mailman/listinfo/opsawg
Attachment:
smime.p7s
Description: S/MIME cryptographic signature