Open Stack

Hi James

Great question and some good answers so far although I think they done
go far enough.
They have (mostly) focused on other operators who, while a very
important part of any cloud, are not the primary audience.
Think more about the end users: devs, qa, marketing, data dudes etc etc.

Focus less on running a cloud and more about how to use the cloud.
As a simple example with huge ramifications, a fully automated and
dynamic build and release pipeline driven by (something like) Jenkins.
1) A commit happens
2) Jenkins created N temporary vm's to run unit tests
3) Tests pass and Jenkins creates more temporary vm's to run integration tests
4) Tests pass and Jenkins creates more temporary vm's to run performance tests
5) Tests pass and Jenkins creates some permanent vm's to add to the
production pool
6) Analytics monitors the response/error rates of the new deployment
7) All looks good so Jenkins is triggered to tear down the production
vm's and replace them (including all Load balancer and networking
rules)
8) Profit!

All this took place using common tools and the Openstack API's. Once
the pipeline is defined, there is little need for human interaction.
Think about ideas like Canary testing, Continuos Deployment, Rapid Prototyping.

Hope thats helpful.

BR,
Stuart

On Tue, Dec 30, 2014 at 7:40 AM, matt <matt at nycresistor.com> wrote:
> There are several fundamental problems in getting your feet wet in
> OpenStack.
>
> First is that OpenStack is expecting to be installed into a rack of systems,
> not one system.  While there are work arounds such as devstack, they fail to
> accurately produce a production environment or even a useful facsimile of
> one.  One of the larger problems for many adopters of openstack at least
> initially is coming to grips with neutron and getting it plugged into their
> existing network environment.  Devstack doesn't help you here much at all.
>
> The other major problem is that there is something of a variety of flavors
> in terms of OpenStack, CFM, Networking, etc.  OpenStack is flexible and
> supported by all the things.  And as a result, there is no simple obvious
> path for deploying it.  Some use puppet, some use chef, some ansible.  Some
> use redhat, some ubuntu, some suse.  Some use cisco, some arista, some
> bigswitch.  The options go on and on.
>
> I think there was a view early on that distributions of openstack would
> emerge.  This has not been the case.  The market has spoken and it has
> rebuked this idea.   Others thought the future would be in appliances.
> Again the market spoke and rebuked this idea.  There is an almost
> pathological hatred of appliances in the enterprise world after all.
>
> So here we are.  Facing a chicken and egg problem.  For the established
> openstack user, you can use one of your greenfield environments to train up
> new employees, and many major deployers offer this training even to outside
> customers.  The problem of course being you are being trained on their
> flavor of openstack, which may not be your own.
>
> But, for the folks who are deploying for the very first time, they are
> entering into a sea of ideas, complex design patterns, and general
> distributed computing theory that is quite literally the subject of many
> masters and phd thesis today.  It's not easy.  And training up to a level of
> certitude in these murky waters is not easy.
>
> I don't have an answer for you, in part because as I've been trying to spell
> out in detail, the scope is enormous.  OpenStack sits upon a cross roads.
> And if you've not found it by following one of those pathes you may need to
> go back and begin the journey again at the beginning.
>
> =/
>
>
> On Tue, Dec 30, 2014 at 7:36 AM, Ritesh Raj Sarraf <rrs at researchut.com>
> wrote:
>>
>> Hello James,
>>
>> On 12/30/2014 02:27 AM, James Dempsey wrote:
>> > So, back to my original question:
>> >
>> > What information do you give new users to help them be effective in the
>> > cloud?  What is your go-to demo for people who don't quite understand
>> > what OpenStack is offering?  How do you reach out to people in your
>> > organizations who aren't OpenStack users yet, but probably should be?
>>
>>
>> Like you, I'm fairly new to OpenStack, or even to the whole concept of
>> Cloud in general.
>>
>> For what I have concluded so far, OpenStack (or Cloud computing in
>> general), is a management interface for all your IT needs.
>>
>> Needs that comprise of Compute, Network and Storage.
>>
>> Cloud computing takes hundreds on physical nodes (comprising of compute,
>> network and storage), makes a database of their abstracted interfaces
>> (virtualization mostly) and provides a unified view.
>>
>> It allows you to spawn of virtual entities anywhere in that farm, thus
>> making it a cloud.
>>
>> Management of this entire infrastructure, is the key component here.
>> That's what OpenStack is.
>>
>>
>> So far, the best start-up guide, for playing around is DevStack.
>> https://github.com/openstack-dev/devstack
>>
>> I don't find it very useful. One for because it is tightly tied to
>> specific releases. And secondly because the entire process is very
>> monolithic.
>>
>>
>> Given the strong management stack that's been designed, for abstracted
>> components, I'd have wished to see support for Linux Containers. That'd
>> have allowed to prototype a 50 node setup on a single dev box.
>>
>> The only thing not covered, would have been the compute component, i.e.
>> afaik Openstack Nova.
>>
>> --
>> Given the large number of mailing lists I follow, I request you to CC me
>> in replies for quicker response
>>
>>
>> _______________________________________________
>> OpenStack-operators mailing list
>> OpenStack-operators at lists.openstack.org
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>
>
>
> _______________________________________________
> OpenStack-operators mailing list
> OpenStack-operators at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>



-- 
BR,
Stuart