Dredging up this thread because I was reminded of it today by a question on ask.openstack.org.

On 18/07/14 09:19, Ayenson, Michael D. wrote:
Hello All,

My name is Mika Ayenson and I have to privilege to intern at Johns Hopkins - Applied 
Physics Lab. I'm really excited to  release the latest proof of concept 
"Re-Heat"  Re-Heat is a JHUAPL developed tool for OpenStack users to help them 
quickly rebuild their OpenStack environments via OpenStack's Heat .

Here is a link to the Re-Heat paper: 
Here is a link to Re-Heat: https://github.com/Mikaayenson/ReHeat

I have included the abstract to our paper here:

This makes me sad. Not because it isn't great work - I'm sure it is. It makes me sad because when I read statements like:

In the context of “entire lifecycle” management, Heat is the “create” aspect of 
OpenStack orchestration.

I realise that we have completely failed to communicate what Heat is :(

To be clear, in the context of "entire lifecycle" management, Heat is the "entire lifecycle" aspect of OpenStack orchestration.

I know I, and I suspect many of us, always hoped that this would be exactly the kind of application where Heat could make a difference, helping scientists to make their research more repeatable.

Heat does that by allowing you to represent your infrastructure as code, and store it under version control. Messing with it behind Heat's back instead of by modifying the template is the infrastructure equivalent of connecting a debugger and messing with the machine code at runtime instead of changing the source. It's the opposite of repeatable. And developing tools to make using this broken pattern more convenient is a step in the wrong direction IMHO.

I strongly recommend you try using the stack update mechanism instead. It's not perfect, but it's getting better all the time. We welcome any feedback you have.

To be clear, I do think there is a really good use of this kind of technology, and it's the one that the Flame developers are targeting: bringing existing applications under Heat's management.



OpenStack has experienced tremendous growth since its initial release just over 
four years ago.  Many of the enhancements, such as the Horizon interface and Heat, 
facilitate making complex network environment deployments in the cloud from scratch 
easier.  The Johns Hopkins University Applied Physics Lab (JHU/APL) has been using 
the OpenStack environment to conduct research, host proofs-of-concepts, and perform 
testing & experimentation.  Our experience reveals that during the environment 
development lifecycle users and network architects are constantly changing the 
environments (stacks) they originally deployed.  Once development has reached a 
point at which experimentation and testing is prudent, scientific methodology 
requires recursive testing be conducted to determine the repetitiveness of the 
phenomena observed.  This requires the same entry point (an identical environment) 
into the testing cycle.  Thus, it was necessary to capture all the changes made to 
the initial !
t during the development phase and modify the original Heat template.  However, OpenStack 
has not had a tool to help automate this process.  In response, JHU/APL developed a 
poof-of-concept automation tool called "Re-Heat," which this paper describes in 
I hope you all enjoy this as I have truly enjoyed playing with HEAT and 
developing Re-Heat.

Mika Ayenson

OpenStack-dev mailing list

OpenStack-dev mailing list

Reply via email to