Hybrid Cloud Computing has two different models:
In the first model, you integrate your on-premises with AWS, Azure or Google, often with the objective to use the storage and dynamic development and deployment capabilities of the Cloud, but keep critical data, systems and code onsite as well. This may be due to regulations, prudence, cost concerns or other business and technological constraints and drivers.
In the second model, many firms have deployed OpenStack or use Rackspace Openstack platforms, Openshift (a better option), or VMWare Tanzu, Kubernetes Hybrid, or the defunct Pivotal Cloud Foundation, to try to imitate the Public cloud characteristics. There are some problems with this:
-Openstack for instance has 15 plus moving components, necessitating expertise in various areas. Kubernetes is likewise a complex set of technologies with limited skills availability in the market.
-Setting up an on-premises imitation of the Public Cloud is expensive, a long term endeavour and will require skills, training and constant upgrades and version control of the components. You are in essence, unwrapping, disaggregating and pulling apart the Public Cloud components and building everything yourself (disassembling your car and then re-assembling).
-PCF and Openstack are incredibly expensive over the longer-term to build, maintain and adapt to changing business requirements
VMWare and Openshift do provide on-site to Cloud integration and are much better choices than Openstack or PCF. Both VMware and Openshift can be integrated with and deployed on Azure and AWS. So, if you are a RedHat shop using RHEL or CentOS, and want to containerise, Openshift hybrid makes sense. If you are a large enterprise and want to standardise on a hypervisor, VMWare provides a best of breed platform. You could virtualise your estate on-premises with VMWare and have a roadmap and glide path into AWS to perform elastic computing.
There is of course a third and for some firms a better model to provide a true on-premises Cloud experience and that is through using the Public Hyperscaler on-premises integration models:
Each of these platforms is essentially an extension of the public cloud offering into on-premises.
AWS Outpost: AWS provides a fully managed service. AWS drops off pre-configured hardware and software to your on-premises data centre or co-location space. As AWS CEO Andy Jassy said when it was introduced: "Customers will order racks with the same hardware AWS uses in all of our regions, with software with AWS services on it."
These services include EC2 instances, EBS storage volumes, Amazon ECS, Amazon EKS clusters for container-based applications and on and on. Outpost providers end users single-digit millisecond latency.
Azure Stack Hub: Can be run on servers from partners, such as HPE, Dell EMC or Lenovo. Microsoft, or one of its partners, can install Azure on the hardware you already own. The objective with Stack is to scale jobs which are below hyper-scale volumes. An example would be to run jobs with 20 VMs on Azure Stack, and other jobs which consume 100s of VMs in Azure Cloud, or run critical but fluctuating and smaller processes on-premises more cost-effectively, than in Azure Cloud.
Google Anthos: Anthos is the production version of Kubernetes cloud which comes out of Google Borg. You can run Anthos on pretty much any standard x86 hardware. It has the same general objectives as Azure Stack, with a focus on Data manipulation and analysis.
IBM Cloud Satellite: This extends it’s the Bluemix control plane to hardware running on-premises. It's a fully vendor-managed hybrid cloud. To entice users IBM offers Cloud Pak for Data and OpenShift as a Service, and it plans to offer third-party software applications on Cloud Satellite.
All 4 of these hybrid offerings are attempting to intersect and promote cloud and edge computing. They are also interested in ‘lock-in’ and controlling a corporation’s migration from on-premises to a targeted and primary Cloud platform. The reality of IT is that there is always and already a ‘lock-in’, though Cloud can provide some degree of portability.