[placement][ptg] Placement Over The Parapet
From the etherpad [1]:
* What's going on in the rest of the world: * where placement can help * that can help placement It's a whole big world out there, people doing all sorts of things that maybe look a bit like what placement does. What can we learn from them? What can they learn from us? In what ways can we expand placement world dominance and improve its ecosystem diversity? We are often so busy in the depths of OpenStack that it is hard to remain aware of what is out there. What is? For example, would there be any purpose in experimenting with plugging placement into the pluggable scheduling concept in k8s? Is that totally crazy or intriguing? I don't know because it's hard to make the time, but it would be valuable to know. [1] https://etherpad.openstack.org/p/placement-ptg-train -- Chris Dent ٩◔̯◔۶ https://anticdent.org/ freenode: cdent tw: @anticdent
On Apr 9, 2019, at 12:08 PM, Chris Dent <cdent+os@anticdent.org> wrote:
It's a whole big world out there, people doing all sorts of things that maybe look a bit like what placement does. What can we learn from them? What can they learn from us? In what ways can we expand placement world dominance and improve its ecosystem diversity?
I think we need to establish world dominance before we plot on how to expand it. :)
We are often so busy in the depths of OpenStack that it is hard to remain aware of what is out there. What is?
I've shared this blog post [0] on the etherpad, but perhaps a wider audience might find it interesting. It caught my eye because last year in Dublin as we started to dig deeper into nested resource providers, the way things fit together reminded me of some of the things I had played around with when I looked into graph databases. I ran a series of experiments that I wrote about in my blog (series beginning with [1]), and am seriously thinking about digging deeper before we get too far down the nested rabbit hole. Even though the blog post was dealing with tracking security groups, it was reassuring to find someone else who also saw cloud resource management as being best modeled as a graph. So maybe this response isn't what you were looking for (how can we integrate Placement into the wilds of the non-OpenStack world?), but rather how can we integrate some of the lessons of the non-OpenStack world into Placement? [0] https://cloudbootup.com/post/cloud-management-with-prolog.html [1] https://blog.leafe.com/hammer-and-nail/ -- Ed Leafe
You might be interested in: https://www.openpolicyagent.org/ in particular Rego (https://www.openpolicyagent.org/docs/v0.10.7/how-do-i-write-policies) Which is inspired by Datalog which is a subset of Prolog So there are other places in the stack reevaluating older tools for use in new places. :) Thanks, Kevin ________________________________________ From: Ed Leafe [ed@leafe.com] Sent: Wednesday, April 10, 2019 12:50 PM To: openstack-discuss@lists.openstack.org Subject: Re: [placement][ptg] Placement Over The Parapet On Apr 9, 2019, at 12:08 PM, Chris Dent <cdent+os@anticdent.org> wrote:
It's a whole big world out there, people doing all sorts of things that maybe look a bit like what placement does. What can we learn from them? What can they learn from us? In what ways can we expand placement world dominance and improve its ecosystem diversity?
I think we need to establish world dominance before we plot on how to expand it. :)
We are often so busy in the depths of OpenStack that it is hard to remain aware of what is out there. What is?
I've shared this blog post [0] on the etherpad, but perhaps a wider audience might find it interesting. It caught my eye because last year in Dublin as we started to dig deeper into nested resource providers, the way things fit together reminded me of some of the things I had played around with when I looked into graph databases. I ran a series of experiments that I wrote about in my blog (series beginning with [1]), and am seriously thinking about digging deeper before we get too far down the nested rabbit hole. Even though the blog post was dealing with tracking security groups, it was reassuring to find someone else who also saw cloud resource management as being best modeled as a graph. So maybe this response isn't what you were looking for (how can we integrate Placement into the wilds of the non-OpenStack world?), but rather how can we integrate some of the lessons of the non-OpenStack world into Placement? [0] https://cloudbootup.com/post/cloud-management-with-prolog.html [1] https://blog.leafe.com/hammer-and-nail/ -- Ed Leafe
On Apr 10, 2019, at 3:11 PM, Fox, Kevin M <Kevin.Fox@pnnl.gov> wrote:
You might be interested in: https://www.openpolicyagent.org/ in particular Rego (https://www.openpolicyagent.org/docs/v0.10.7/how-do-i-write-policies) Which is inspired by Datalog which is a subset of Prolog
That does seem to be an interesting approach.
So there are other places in the stack reevaluating older tools for use in new places. :)
Oh, geez, I wasn’t hoping to start writing Prolog code! :) -- Ed Leafe
On Tue, Apr 9, 2019 at 12:09 PM Chris Dent <cdent+os@anticdent.org> wrote:
For example, would there be any purpose in experimenting with plugging placement into the pluggable scheduling concept in k8s? Is that totally crazy or intriguing?
StarlingX has on its futures list being able to share a compute node between k8s and Nova and it seems to me as though placement is the heart of doing that, so totally not crazy. I am thus far ignorant enough about k8s as to not be able to reliably judge on technical merits but would love to further discuss. dt -- Dean Troyer dtroyer@gmail.com
On Thu, Apr 11, 2019 at 5:55 PM Dean Troyer <dtroyer@gmail.com> wrote:
On Tue, Apr 9, 2019 at 12:09 PM Chris Dent <cdent+os@anticdent.org> wrote:
For example, would there be any purpose in experimenting with plugging placement into the pluggable scheduling concept in k8s? Is that totally crazy or intriguing?
StarlingX has on its futures list being able to share a compute node between k8s and Nova and it seems to me as though placement is the heart of doing that, so totally not crazy. I am thus far ignorant enough about k8s as to not be able to reliably judge on technical merits but would love to further discuss.
I think Zun could benefit from this too, to share compute resources across different services.
dt
-- Dean Troyer dtroyer@gmail.com
-- Mohammed Naser — vexxhost ----------------------------------------------------- D. 514-316-8872 D. 800-910-1726 ext. 200 E. mnaser@vexxhost.com W. http://vexxhost.com
participants (5)
-
Chris Dent
-
Dean Troyer
-
Ed Leafe
-
Fox, Kevin M
-
Mohammed Naser