[Container-tools] Nulecule Application Story & examples - illustrations

[Mark Lamourine]

----- Original Message -----
> 
> 
> Sent from my iPhone
> 
> > On May 7, 2015, at 2:38 PM, Mark Lamourine <markllama at redhat.com> wrote:
> > 
> > 
> > 
> > ----- Original Message -----
> >> 
> >> Hi,
> >> 
> >> Draft illustrations - Nulecule Application Story & examples:
> >> https://docs.google.com/presentation/d/1rRfR9sSLuz_hMmz6W5SZAsV2AKDgHmfJk0hhFF0TTjU/edit#slide=id.p
> >> 
> >> Any comments, suggestions, or corrections? Did I get the examples right?
> > 
> > I look at Docker/Rocket, Kubernetes Mesos and Nulecule as russian dolls.
> > (mostly)
> > 
> > The container system creates a tiny capsule environment on a host.  It has
> > holes or sockets where the outside world is able to/required to connect to
> > feed in resources: network, files, environment.  Each container runs only
> > on one host.  The container system's scope is on the host and with very
> > few exceptions it is unable to make changes to the configuration of the
> > host itself.  The unit of operation for the container system is one
> > container.  It can create containers that share resources but it cannot
> > *provide* the resources if they are not already available on the host.
> > 
> > The orchestration system has two functions.  It prepares the host to start
> > a container by rallying the external resources the container will need.
> > It finds and attaches the storage, it creates and publishes network
> > interfaces.  It also binds multiple hosts together, examining the
> > resources and loading of each one so that it can report that to a
> > scheduler and then place and execute each container as it is instructed.
> > Because the orchestration system is aware of multiple running containers
> > and has control of resources they might share.  It can modify the host to
> > provide those resources to multiple local containers, or even containers
> > which span hosts in the cluster.
> > 
> > The scheduler (mesos) takes information about the resources available in
> > the cluster (from Kubernetes) and a provided policy from creators of the
> > cluster and the jobs from a user queue.  It matches the jobs to the
> > resources by applying the policy and then hands the jobs down to the
> > orchestration system to execute.
> > Mesos uses the orchestration system as its hands.
> > 
> > Nulecule is the language in which one describes a complex running service
> > in the most abstract terms.  It provides input to both the orchestration
> > and scheduling systems.  It is how you say "I want to do *this*" and not
> > "turn that knob, plug that cord".
> > 
> > Some of the details will need to be built in like plugins to span the
> > layers.  Aaron mentioned to me today that there were hooks already in
> > nulecule to have "provider" plugins, a familiar concept to Puppet users,
> > so that each abstracted concept has a concreted implementation for each
> > environment.
> > 
> > All this presumes a selection of resource legos, container images, storage,
> > external network ports etc.
> > 
> > Does this make any sense?
> 
> Yes but I don't know how you crept Mesos in here. I didn't know it was in
> scope. I thought we were still scratching our heads over that.

I'm thinking conceptually, not just architecturally.  Without a scheduler in there it's still not "automated". It's a stretch, not a plan.

> 
> William
> > 
> > - Mark
> > 
> > 
> >> 
> >> Thanks.
> >> Michael
> >> 
> >> _______________________________________________
> >> Container-tools mailing list
> >> Container-tools at redhat.com
> >> https://www.redhat.com/mailman/listinfo/container-tools
> > 
> > --
> > Mark Lamourine <mlamouri at redhat.com>
> > Sr. Software Developer, Cloud Strategy
> > Red Hat, 314 Littleton Road, Westford MA 01886
> > Voice: +1 978 392 1093
> > http://people.redhat.com/~mlamouri
> > markllama @ irc://irc.freenod.org*lopsa
> > 
> > _______________________________________________
> > Container-tools mailing list
> > Container-tools at redhat.com
> > https://www.redhat.com/mailman/listinfo/container-tools
> 

-- 
Mark Lamourine <mlamouri at redhat.com>
Sr. Software Developer, Cloud Strategy
Red Hat, 314 Littleton Road, Westford MA 01886
Voice: +1 978 392 1093
http://people.redhat.com/~mlamouri
markllama @ irc://irc.freenod.org*lopsa