openstack team mailing list archive

[Lorin Hochstein <lorin@xxxxxxx>]

Hi Sandy:

Do you know if there are there any nova branches floating out there that implement Capabilities? I've seen the following related wiki pages and blueprints, but I haven't found any implementations out there yet.

https://blueprints.launchpad.net/nova/+spec/extra-data
https://blueprints.launchpad.net/nova/+spec/use-metadata-tags-for-capabilities
http://wiki.openstack.org/RequestCapabilities

The zone scheduler slides mention capabilities as well, but those are at the "zone" level, and we're potentially interested in compute nodes with different capabilities within the same zone. 

(Note that for our use case, we'll probably be using flavors to specify different capabilities, so we don't (yet) need the users to be able to send arbitrary strings using the API). 

Lorin
--
Lorin Hochstein, Computer Scientist
USC Information Sciences Institute
703.812.3710
http://www.east.isi.edu/~lorin

On May 11, 2011, at 11:54 AM, Sandy Walsh wrote:

> Hi Lorin,
> 
> Zones can have multiple parents. Child Zones don't know their parents ... parents only know about children.
> 
> Sharing services across Zones isn't permitted (since they would need to share a DB and AMQP). 
> 
> You could solve the problem by using the Capabilities to determine where instances are created (GPU only or not) or create smaller sub-zones for GPU/not-GPU and let the parent zones select between them. 
> 
> -S
> 
> 
> From: openstack-bounces+sandy.walsh=rackspace.com@xxxxxxxxxxxxxxxxxxx [openstack-bounces+sandy.walsh=rackspace.com@xxxxxxxxxxxxxxxxxxx] on behalf of Lorin Hochstein [lorin@xxxxxxx]
> Sent: Wednesday, May 11, 2011 12:06 PM
> To: Openstack
> Subject: [Openstack] Zones / distributed scheduler question
> 
> All:
> 
> For the proposed zones / distributed scheduler functionality <https://blueprints.launchpad.net/nova/+spec/zones101>, can we defines  zones that overlap without being nested? 
> 
> Consider the scenario where: 
> 
> * We have two datacenters: one in Arlington, Virginia (ARL), and one in Marina del Rey, California (MDR).
> * At each datacenter, we have both conventional x86 nodes and  nodes with GPUs 
> 
> Can we define our zones such that:
> 
> * We have a zone for each datacenter (ARL, MDR)
> * We have a zone for GPU nodes  (GPU)
> 
> In this case, the GPU zone would both overlap the ARL and MDR zones, but ARL and MDR would not be fully nested within GPU. 
> 
> 
> Lorin
> --
> Lorin Hochstein, Computer Scientist
> USC Information Sciences Institute
> 703.812.3710
> http://www.east.isi.edu/~lorin
> 
> Confidentiality Notice: This e-mail message (including any attached or
> embedded documents) is intended for the exclusive and confidential use of the
> individual or entity to which this message is addressed, and unless otherwise
> expressly indicated, is confidential and privileged information of Rackspace.
> Any dissemination, distribution or copying of the enclosed material is prohibited.
> If you receive this transmission in error, please notify us immediately by e-mail
> at abuse@xxxxxxxxxxxxx, and delete the original message.
> Your cooperation is appreciated.