A branching-out suggestion: virt-manager for cloud

Matt Wagner matt.wagner at redhat.com
Fri Apr 5 20:30:13 UTC 2013


Hi folks,

I was chatting a bit with Nitesh earlier today. (He wrote the 
Aeolus-gui[1] app a while back, as you may recall.)

First off, it made me realize that the current situation is probably 
unclear and alienating to people who aren't working on Aeolus full-time. 
We've been talking a bit about whether we want to move away from 
Conductor and in favor of a cloud broker approach, and then we've also 
been talking about leaving Fedora, etc. I think we're moving in the 
right direction and Aeolus will be a better project because of it, but 
I'm not sure things have been made abundantly clear on the list. (Though 
the wiki page helps.)

We were chatting a bit about what would be beneficial to the Aeolus 
project as a whole, and floated the idea of some sort of lightweight 
little app for managing cloud instances. And an idea sort of formed.

I think it would be interesting to build something like virt-manager[2], 
but for managing cloud instances. Nothing fancy, and probably not 
scalable to the enterprise market where people might be managing a 
handful of instances.

It could just be a GUI client for Deltacloud. Maybe Qt, as Nitesh's 
specialty.

I think there are a few reasons this could be pretty neat:

1.) It's easy to start small and scale up. Grab a list of instances from 
Deltacloud, and let the user pick one to launch, on a given hardware 
profile, and show a list of running instances. Over time, you can work 
in reporting on statistics; disk/network/etc. management, etc.; a VNC 
client where supported.

2.) You could support multiple providers the same way that virt-manager 
supports multiple hypervisors -- just add multiple, and show them 
separately. You can have a collapsible EC2 section and a collapsible 
OpenStack section in the same app.

3.) It could be a fun project while we work on figuring out what we're 
doing long-term.

4.) Most of our plans for an eventual cloud broker involve exposing a 
Deltacloud API in Aeolus, that would transparently map to the ideal 
cloud provider. Thus, this app could work with whatever we end up building.

5.) Having a lightweight client for whatever we're building would be 
pretty useful, so we wouldn't be constrained to the API-only. And having 
it be a generic client for clouds would emphasize our value-add, versus 
being a heavy monolithic app.

6.) It might attract a different crowd than "enterprise hybrid cloud" 
attracts. I think that's a good thing, especially upstream. Enterprise 
hybrid cloud management is sort of a niche market. I think there are 
many more people that would be interested in a lightweight little 
desktop GUI for managing cloud instances, but that some of the problems 
we'd be solving overlap a lot. IOW, I think it has potential to grow the 
community, and that this would still translate into building a better, 
more robust enterprise-y management app.

What do you think? Have I gone insane? Would this be a worthwhile endeavor?

-- Matt


[1] https://github.com/niteshnarayanlal/Aeolus-gui
[2] http://virt-manager.org/


More information about the aeolus-devel mailing list