- Developer Tools
Troy Topnik, April 17, 2014
When companies move from bare metal to virtualization, it's an enormous step that promises cost savings and greater efficiency (which it usually delivers). But with the virtualization step complete, some companies still struggle with building out an Infrastructure-as-a-Service layer (IaaS). Virtualization is really only the first step in the process - the IaaS includes other components such as the provisioning and management not only of the VMs, but also networks and storage.
However, a lot of people we talk to are ready to deploy their cloud apps, but feel that their IaaS needs to be complete and ready to go before they can move up the cloud stack with a PaaS product like Stackato.
The reality is that they can actually start down the path of "moving to the cloud" without having a fully functional on-prem IaaS in place. With Stackato you can actually get started on building your cloud stack without needing the IaaS layer right away.
Even if you're still defining and strategizing what your IaaS layer should be, you can set up Stackato on a variety of hypervisors (KVM, VirtualBox, vSphere) and start getting the benefits of using an application platform to deploy your apps (like faster time to market, easier management of applications, delivering business value faster, etc.). In a prior blog Ho Ming Li, our Cloud Solution Architect, set up a cluster using a few of the hypervisors that Stackato supports. While this is not a recommended configuration, it shows how simple it can be to install Stackato without a homogenous private cloud infrastructure in place.
Once we've proven to a prospect that Stackato works on their hypervisor, the next most commonly asked question we get is, "What happens when we've finalized our IaaS?" Again, no problem. The infrastructure-agnostic nature of Stackato allows you to plug right into whatever cloud infrastructure you ultimately build. Stackato is ready to work on OpenStack, CloudStack, vSphere, and Eucalyptus.
A PaaS-First Approach
Stackato is distributed as a virtual machine. To make a cluster, you start up a number of these machines on whichever hypervisor you have handy, then link them together and assign roles to each VM using a couple of simple commands.
These Stackato VMs are now nodes in a Stackato cluster, ready for users and applications.
Multi-tenancy (user accounts, organizational groupings, quotas, etc.) is already part of Stackato, so the infrastructure layer does not necessarily have to have those features. End user applications consume containers and service instances rather than virtual machines. The people deploying and managing applications on the PaaS do not need access to the hypervisor or IaaS layer.
At larger scales, where a company is subdividing large virtualized server farms, IaaS features become much more important. But for initial deployments, especially proof-of-concept systems, running a cluster of one or two dozen VMs is manageable for most IT admins who are familiar with their hypervisor.
In this approach, you essentially skip the "as-a-service" part of IaaS and deal more directly with your infrastructure. The Stackato admins in this scenario would probably be provisioning virtual machines on hardware they have some kind of access to.
That's fine. Not everyone has a private cloud at their disposal, and sometimes the person who spins up the VMs is the same one who sorts out the cables in the rack.
A Migration Path to Larger Scale
What happens when the project takes off? If the proof-of-concept PaaS is a success, and usage of the system outgrows what the rack in your co-lo can provide? Well, there are simple ways to move an entire Stackato cluster, with all the apps and data, to new locations or platforms.
Before the 3.2 release, major version upgrades of Stackato were done by starting up VMs of the newer version and migrating data to them from the old cluster. That's no longer necessary for upgrades, but the same migration mechanism is available for moving Stackato wherever you need it to go. The import/export commands don't care about the underlying location, hardware, or hypervisor of your clusters. If you want to move your PaaS to a public cloud provider, that's fine too. As long as you can boot Stackato VMs, you can move your existing data there.
Start Small and Grow With Adoption
Since companies tend to move to the cloud in stages, by using Stackato for the PaaS layer they can start small and still reap the benefits of using an application platform with the knowledge that their PaaS will scale with their project.
So, no IaaS? No problem. Download the Stackato VM and get started.
Subscribe to ActiveState Blogs by Email
Share this post: