Open Stack

On 04/29/2013 05:25 AM, Thierry Carrez wrote:
> Monty Taylor wrote:
>> I'm actually less interested in this as a generic PaaS on top of
>> Heat/OpenStack. That might be because I don't see DBaaS as a PaaS thing
>> at all, and also because I don't really want to go down that rathole at
>> the moment.
>>
>> Databases are clearly something that people need for applications, and
>> that are hard to administer properly. PaaS solutions are a thing that
>> while useful to a set of people, does seem like a scope creep to me for
>> the project, and one where a general framework provided by us seems
>> premature.
> 
> I'd agree with that. Even if databases are way higher in the OSI stack,
> I see databases as a basic infrastructure building block in the same way
> as load balancers. In my view there would be three types of IaaS/PaaS users:
> 
> 1. Pure players that only consume extremely raw (some would say "core")
> resources: VMs, block storage, basic networking -- and build everything
> else themselves on top of that
> 
> 2. Application builders which benefit from using building blocks for
> almost-always-used functions: DNS, load balancing, databases, message
> queues...
> 
> 3. One-stop deployers who just push a repo and the PaaS platform handles
> deployment/scaling and the details of their infrastructure for them --
> Google AppEngine or Heroku style consumers
> 
> (1) is clear IaaS, (3) is clear PaaS. (2) is somewhere in between. Let's
> call it IaaS++. You still care about the details of your infrastructure
> and which raw resources you consume, but you enjoy a number of
> higher-level services. You could build those services yourself on top of
> "core IaaS" services, but that's a pain to setup and maintain, so you
> enjoy the convenience of IaaS++ services.
> 
> You could argue that the only things in "core IaaS" are VMs, block
> storage and raw networking, since you can build everything else on top
> of that. You could argue that object storage is actually part of IaaS++.
> 
> At this point, I think considering IaaS++ projects as part of the
> OpenStack integrated release is completely valid, since we already
> started to enter that space. I would certainly be more hesitant with
> projects that belong in (3).

Well said. Totally agree.