[Isod-rg] OGF ISOD-RG use case document

[Joan A. Garcia-Espin]

Hi all,

I'm glad to see discussion here!
My view is that generic infrastructure services for the cloud cannot  
be limited to the commonly known IaaS/PaaS/SaaS concepts. *aaS in the  
IT world don't really mind about the network, or it does using the  
maxim "bandwidth is cheap". Hence, we should discuss how to  
incorporate the 'network' into these paradigms. As an example, the  
work done in the NSI-wg can serve as the basis for dynamic network  
infrastructure services for the cloud.

Best regards,
--
Joan A. García-Espín
DANA, Fundació i2CAT







El 16/09/2011, a las 18:49, Yuri Demchenko escribió:

> Hello Everybody,
>
> Some comments inline. Apologies for joining a bit late to the thread.
>
> On 16/09/2011 17:34, Alexander Willner wrote:
>>>>> to PaaS
>>>> What about "nPaaS" (Network PaaS): here you can specify network  
>>>> properties
>>>> of your Application PaaS.
>>> My idea is: ok, a network domain has some properties (if managed)  
>>> that
>>> allows you to provide net services on top. Hence, one can see the  
>>> net
>>> domain as a "platform" for net services. This has implications in
>>> the management and control functions of the domain: you need some  
>>> generic
>>> functions, and some tech-specific ones. Does this reasoning fit  
>>> your mind?
>>
> Current Cloud concept and models take network as ubiquitous/ 
> omnipresent
> property of the environment in which Cloud service models are  
> realised.
>
> But we all know that in practice many cases require that network  
> quality
> is guaranteed, or provisioned as part of requested service with
> specified parameters.
>
> I think this is one of our goal to define and justify need for
> guaranteed network QoS in Cloud provisioning.
>
>> I think we can agree the e.g. Google App Engine is a typical PaaS  
>> example.
>> Here, on this layer, we should incorporate the network. So we don't  
>> want
>> to setup a network topology or configure interfaces. As you wrote:  
>> "you
>> need some generic functions". What if I can deploy a "network-aware"
>> distributed application to the Google cloud? I.e. I can specify that
>> I want the network e.g. behave like GPRS mobile network. Is this  
>> what we
>> talk about?
>>
> But for both service models PaaS and SaaS network is not a part of  
> this
> model.
>
> So, we could either motivate special brand of network aware PaaS/SaaS
> Cloud services e.g. PaaS-N and SaaS-N (I am not using proposed acronym
> nPaaS because it is different :-), or attribute provisioning  
> guaranteed
> network services to a 3rd party - Network Carrier (as defined in the
> recent NIST documents [1]), or using GEYSERS approach to Cloud
> Operator/Integrator.
>
>>>>> and, why not, to SaaS.
>>>> Mh, Network as a Software?
>>> Not this way, I am referring to "network software". Simple  
>>> example: the control plane and its behaviour. Complex example:  
>>> programmable networks. It is again related to control and  
>>> management, but now in thinking in the operation phase. Even more,  
>>> using the Google App Engine (one platform) one can run different  
>>> apps simultaneously (several sw pieces). Is this applicable to  
>>> networking? How do we re-define SaaS if so?
>>
>> What about this:
>>
>> a) I can setup a virtual NSI/Harmony/Argon/Argia/... instance  
>> somewhere
>> in the cloud that can control the physical/virtual network I'm  
>> managing or
>> b) a can rent a pre-configured appliance that runs an NSI instance
>> for my network...
>>
> This would my understanding of a correct design but you cannot dictate
> this to the Cloud provider, however as I mentioned above this can be  
> in
> the scope and interest of so-called Cloud Operator, Broker or  
> integrator.
>
> This may be a use case which we can propose and analyse.
>
> Yuri
>
> [1]
> http://collaborate.nist.gov/twiki-cloud-computing/pub/CloudComputing/StandardsRoadmap/NIST_CCSRWG_092_NIST_SP_500-291_Jul5.pdf
>
> _______________________________________________
> Isod-rg mailing list
> Isod-rg at ogf.org
> http://www.ogf.org/mailman/listinfo/isod-rg
>