Step one is to follow the lead of T106475: Evaluate a 'cluster solution' for use on Tool Labs and develop evaluation criteria. What features are must haves? What are desirable? These should cover both operational and end user concerns. Use T136265: Develop evaluation criteria for comparing Platform as a Service (PaaS) solutions to discuss your opinions.

Step two involves evaluation, so we need to come up with a list of possible candidates. At the moment I'm aware of Deis and OpenShift. Are there others?

I would like to propose deploying a testing instance not only for the top solution, but (at least) the top two. Say a rough, but working installation takes two weeks engineer time each and a couple of out-of-warranty machines, then that's still a price worth to pay for an informed choice between multiple options that will need to last for several years.

In T136264#2328997, @bd808 wrote:

Step two involves evaluation, so we need to come up with a list of possible candidates. At the moment I'm aware of Deis and OpenShift. Are there others?

There's Flynn and Dokku. Shipyard might be appropriate too.

In T136264#2330394, @scfc wrote:

I would like to propose deploying a testing instance not only for the top solution, but (at least) the top two. Say a rough, but working installation takes two weeks engineer time each and a couple of out-of-warranty machines, then that's still a price worth to pay for an informed choice between multiple options that will need to last for several years.

I agree, yeah. When we evaluated kubernetes vs mesos/marathon we setup test installations of both of them so we can do a handson evaluation, and I think we'll have to do the same here.

In T136264#2330485, @tom29739 wrote:

In T136264#2328997, @bd808 wrote:

There's Flynn and Dokku. Shipyard might be appropriate too.

We should only consider things that run on top of Kubernetes, IMO.

In T136264#2331069, @yuvipanda wrote:

In T136264#2330485, @tom29739 wrote:

In T136264#2328997, @bd808 wrote:

There's Flynn and Dokku. Shipyard might be appropriate too.

We should only consider things that run on top of Kubernetes, IMO.

Oh, sorry, I didn't realise that.

http://www.kelproject.com/ is a super new one.

In T136264#2331065, @yuvipanda wrote:

In T136264#2330394, @scfc wrote:

I would like to propose deploying a testing instance not only for the top solution, but (at least) the top two. Say a rough, but working installation takes two weeks engineer time each and a couple of out-of-warranty machines, then that's still a price worth to pay for an informed choice between multiple options that will need to last for several years.

I agree, yeah. When we evaluated kubernetes vs mesos/marathon we setup test installations of both of them so we can do a handson evaluation, and I think we'll have to do the same here.

Summary steps updated to "Deploy testing instance of top two solutions based on evaluation for head to head testing".