[one-ecosystem] Fwd: About features 201 and multi-site management

florian chazal florianchazal at gmail.com
Wed Jun 2 08:33:45 PDT 2010 

---------- Forwarded message ----------
From: florian chazal <florianchazal at gmail.com>
Date: 2010/6/2
Subject: Re: [one-ecosystem] Fwd: About features 201 and multi-site
management
To: Carlos Martín Sánchez <cmartins at fdi.ucm.es>


Hi Carlos,

I finally choose to develop the same kind of architecture as Eucalyptus,
that is to say with an open nebula in each site (Site controller) and one
above all of them (Clouds controller) in order to integrate the location
characteristic. I guess that I have to think about few thing such as :
- How the Cloud Controller will speak with the site controller (OCCI, the
VEEM-VEEM driver from RESERVOIR, one of the APIs JAVA, Ruby ...??)
- May I can develop the same kind of driver that the EC2 one?,
- Integration of the site monitored information and how to get back this
information,
...

Therefore my question is how long do you think I need to develop a driver
like the EC2 to enable the control of my site controller by the cloud
controller?

Thanks in advance,

Florian CHAZAL


2010/6/1 Carlos Martín Sánchez <cmartins at fdi.ucm.es>

Hi Florian and all.
>
> I've discussed with my colleagues the possibility of improving and
> releasing this plug-in as an ecosystem component.
> But right now the only advantages this plug-in may bring you over other
> interfaces depend on the use of other Reservoir components that aren't
> available for download.
>
> You could achieve the same ONE to ONE federation behaviour using the EC2
> interfaces [1] [2]. Take a look into that, and let us know if that solution
> meets your requirements.
>
> Best regards,
> Carlos
>
> [1] EC2 Driver   http://opennebula.org/documentation:rel1.4:ec2g
> [2] EC2 Query service  http://opennebula.org/documentation:rel1.4:ec2qug
>
>
> Carlos Martín, Cloud Technology Engineer/Researcher
> DSA Research Group: web http://dsa-research.org and blog
> http://blog.dsa-research.org
> OpenNebula Open Source Toolkit for Cloud Computing:
> http://www.OpenNebula.org
>
>
> El 1 de junio de 2010 10:34, florian chazal <florianchazal at gmail.com>escribió:
>
> Hi,
>>
>> For the moment I didn't choose which architecture I will use yet (files
>> attached in my previous mail). But I am very  interested to modify and test
>> this VMI plug-in. I have some contacts with the Thales RESERVOIR team in
>> Paris, therefore I will may be have some help from them.
>> Thank you for sending me the plug-in.  I can keep you in touch of my
>> progress  if you are interested in.
>>
>> Regards
>>
>> Florian CHAZAL
>>
>> 2010/5/31 Carlos Martín Sánchez <cmartins at fdi.ucm.es>
>>
>> Hi Florian,
>>>
>>> I could send you the VMI plug-in code if you are interested, but the code
>>> isn't pretty and the current version works with a Reservoir site on the
>>> other end, acting as the OCCI server.
>>> It can be modified to work with another OpenNebula, but could take some
>>> time.
>>>
>>> What the VMI plug-in basically does is add a Reservoir site (or
>>> OpenNebula front-end) as a new regular host; but the target site is seen as
>>> a big machine with no capacity monitoring.
>>>
>>> Other challenging issues would be networking and migration.
>>> A VPN could be established, but doing this automatically isn't trivial.
>>> Technically, migration would work is you set up shared storage across
>>> sites, but this may bring a performance drop.
>>>
>>> Best Regards,
>>> Carlos
>>>
>>> Carlos Martín, Cloud Technology Engineer/Researcher
>>>
>>> DSA Research Group: web http://dsa-research.org and blog
>>> http://blog.dsa-research.org
>>> OpenNebula Open Source Toolkit for Cloud Computing:
>>> http://www.OpenNebula.org
>>>
>>>
>>> 2010/5/31 Ignacio Martin Llorente <llorente at dacya.ucm.es>
>>>
>>>> Hi
>>>>
>>>> >
>>>> > But I don't understand what the OpenNebula instance see ?
>>>> > How does it can submit VM without knowing anything about the targeted
>>>> private cloud?
>>>>
>>>> It manages remote ONE clouds as public clouds, such as Amazon EC2. Good
>>>> enough for loosely coupled federation, but for tightly coupled you would
>>>> need more details (see section 4.1 in
>>>> http://www.reservoir-fp7.eu/uploads/Deliverables/HighLevelArchY2.pdf) ,
>>>> for example regarding the underlying physical infrastructure. I understand,
>>>> this is the part you are interested in.
>>>>
>>>> >
>>>> > 2. Use the hybrid cloud computing functionality provided by ONE to
>>>> access remote cloud sites (this is the idea represented in
>>>> http://claudia.morfeo-project.org/wiki/index.php/Image:RESERVOIRArchitecture.pngand followed by the RESERVOIR architecture).
>>>> >
>>>> > Does it mean that a "VEEM-VEEM" driver has already been done as the
>>>> RESERVOIR web site seems to say
>>>> > (http://62.149.240.97/index.php?page=vee-management, last line) ?
>>>> > And is it the same kind of driver as the EC2 one? If yes you see at
>>>> least the free cpu and free memory available like the targeted cloud is only
>>>> one big machine.
>>>>
>>>> Yes, but this plugin is not available for download, because it is only a
>>>> prototype.
>>>>
>>>> >
>>>> > > No, VPN enables you to have virtual networks across sites but not to
>>>> migrate between sites. You would need a shared file system for this.
>>>> >
>>>> > Then if I set up a shared file system across sites, live migration
>>>> will be possible, isn't it? But is there any solution to avoid shared file
>>>> systems?
>>>>
>>>> This is an issue being addressed by IBM within RESERVOIR, but it is not
>>>> available for download....
>>>>
>>>> Regards
>>>>
>>>> Ignacio
>>>>
>>>>
>>>> >
>>>> >
>>>> > Thanks a lot.
>>>> >
>>>> > Best regards,
>>>> >
>>>> > Florian CHAZAL
>>>> >
>>>> > 2010/5/31 Ignacio Martin Llorente <llorente at dacya.ucm.es>
>>>> > >
>>>> > >
>>>> > > Hi
>>>> > >
>>>> > >> Hi,
>>>> > >>
>>>> > >> Thank you for your useful answer. I don't know if you are involved
>>>> in the RESERVOIR Project
>>>> > >
>>>> > > I am coordinating one its main three technical activities.
>>>> > >
>>>> > >
>>>> > >> but I realize that they seem to have already integrated the idea of
>>>> multiple site (
>>>> http://claudia.morfeo-project.org/wiki/index.php/Image:RESERVOIRArchitecture.png
>>>> ).
>>>> > >
>>>> > >
>>>> > > This is a third approach for managing multiple cluster/sites, "cloud
>>>> brokering", so summarizing:
>>>> > >
>>>> > > 1. Having a single ONE instance to manage several clusters (this is
>>>> the functionality being implemented 201)
>>>> > > 2. Use the hybrid cloud computing functionality provided by ONE to
>>>> access remote cloud sites (this is the idea represented in
>>>> http://claudia.morfeo-project.org/wiki/index.php/Image:RESERVOIRArchitecture.pngand followed by the RESERVOIR architecture). You have your own private cloud
>>>> an on-demand access to remote instances to meet peak demands.
>>>> > > 3. Use a "cloud broker" on top of several cloud sites, you could use
>>>> OpenNebula only as Cloud broker to access remote clouds that could be also
>>>> running OpenNebula.
>>>> > >
>>>> > >
>>>> > >> And we can see that the different site are managed by different
>>>> OpenNebula instance and they communicate by VMI.
>>>> > >
>>>> > > Approach numer 2 above.
>>>> > >
>>>> > >> Do you have more information about that? For instance, Is it
>>>> possible to live migrate a VM between two sites using the VMI (assuming that
>>>> we set up a VPN between site to have the same subnetwork)?
>>>> > >
>>>> > > No, VPN enables you to have virtual networks across sites but not to
>>>> migrate between sites. You would need a shared file system for this.
>>>> > >
>>>> > >> How the different OpenNebula instances are awre about the existence
>>>> of the others?
>>>> > >
>>>> > > They are fully independent administrative domains. A given instance
>>>> does not any internal details of the rest of instances, it is only able to
>>>> submit VMs.
>>>> > >
>>>> > >>
>>>> > >> And if you have any opinion about my diagrams attached you're
>>>> welcome.
>>>> > >
>>>> > > I see that your first diagram represents the third approach (one ONE
>>>> instance as broker of other instances having details form underlying cloud
>>>> instances. You would have to develop new information plugins to collect the
>>>> required details ) and the second one represents the first approach (one ONE
>>>> instance managing several clusters with the new functionality from 201).
>>>> > >
>>>> > > Regards
>>>> > >
>>>> > >>
>>>> > >> Thanks a lot.
>>>> > >>
>>>> > >> Regards,
>>>> > >>
>>>> > >> Florian CHAZAL
>>>> > >>
>>>> > >>
>>>> > >> 2010/5/28 Ignacio Martin Llorente <llorente at dacya.ucm.es>
>>>> > >> Hi
>>>> > >>
>>>> > >> The new functionality in issue 201 will allow users to define the
>>>> cluster where they would like to run the service (group of VMs). Some of our
>>>> users requested this functionality for example to manage clusters with
>>>> different levels of security, or clusters with different hardware devices
>>>> for different types of workloads profiles. It could be also used to define
>>>> geographical location if you have clusters located at different sites.
>>>>  However, the management of different sites is usually performed by
>>>> different OpenNebula instances. So you could create a federation of cloud
>>>> sites by using the support for hybrid cloud computing given by OpenNebula.
>>>> > >>
>>>> > >> Regards
>>>> > >>
>>>> > >>
>>>> > >> --
>>>> > >> Ignacio M. Llorente, Full Professor (Catedratico):
>>>> http://dsa-research.org/llorente
>>>> > >> DSA Research Group:  web http://dsa-research.org and blog
>>>> http://blog.dsa-research.org
>>>> > >> OpenNebula Open Source Toolkit for Cloud Computing:
>>>> http://www.OpenNebula.org
>>>> > >> RESERVOIR European Project in Cloud Computing:
>>>> http://www.reservoir-fp7.eu
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >> On 28/05/2010, at 14:47, florian chazal wrote:
>>>> > >>
>>>> > >>> Hi,
>>>> > >>>
>>>> > >>> I am not sure I am in the good mailing list for this but :
>>>> > >>>
>>>> > >>> I would like to extend the functionality of open nebula by adding
>>>> the geographical location of VMs or (group of VMs as Virtual datacenter)
>>>> with the idea of adding a level above hosts : DataCenters location. Thanks
>>>> to that the client will have the possibility to choose the location of his
>>>> data, like the availability zone in Eucalyptus/EC2. Have you seen this
>>>> feature also with this kind of use case?
>>>> > >>>
>>>> > >>> Thanks!
>>>> > >>>
>>>> > >>> --
>>>> > >>> Florian Chazal
>>>> > >>> _______________________________________________
>>>> > >>> Ecosystem mailing list
>>>> > >>> Ecosystem at lists.opennebula.org
>>>> > >>> http://lists.opennebula.org/listinfo.cgi/ecosystem-opennebula.org
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >>
>>>> > >> --
>>>> > >> Florian Chazal
>>>> > >> <Hierarchical_architecture.jpeg><Horizontal_architecture.jpeg>
>>>> > >
>>>> >
>>>> > _______________________________________________
>>>> > Ecosystem mailing list
>>>> > Ecosystem at lists.opennebula.org
>>>> > http://lists.opennebula.org/listinfo.cgi/ecosystem-opennebula.org
>>>> >
>>>> >
>>>> >
>>>> > --
>>>> > Florian Chazal
>>>>
>>>> _______________________________________________
>>>> Ecosystem mailing list
>>>> Ecosystem at lists.opennebula.org
>>>> http://lists.opennebula.org/listinfo.cgi/ecosystem-opennebula.org
>>>>
>>>
>>>
>>
>>
>> --
>> Florian Chazal
>>
>
>


-- 
Florian Chazal



-- 
Florian Chazal
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opennebula.org/pipermail/ecosystem-opennebula.org/attachments/20100602/2ad99330/attachment-0002.htm>

More information about the Ecosystem mailing list