Selecting the right hybrid cloud deployment model

Hybrid cloud deployments have demonstrated advantages in terms of better economics, scalability and business agility. The ‘best-of-both worlds' approach of hybrid cloud provides the benefits of public clouds of lowering costs by using on-demand scale and pay as you use model while having complete control on the critical data and processes inside the organization, as with private clouds.

It's definitely not an easy task to be able to select the workloads, identify challenges associated with it, and addressing those for a smoother transition to a Hybrid Cloud. Let us first understand the types of integration workloads which are associated with implementing Hybrid Cloud. The types of Hybrid Cloud workloads are categorized based on the level of integration between public and private cloud services.

The next section captures the possible hybrid cloud deployment models based on the integration styles, at Infrastructure Level, Application Level and Data Level. We will try to capture few viewpoints for each deployment model.

Hybrid Infrastructure

This hybrid cloud model works at the Infrastructure level, where the private and public cloud share or extend the virtual machines, and it is mostly done through the virtual machine management and movements.

Work load Selection: This type of hybrid cloud model is suitable when the workload needs on-demand scalable infrastructure as temporary or extension to the on-premise private cloud. Few use cases might be the development environments and variety of test environments and low capacity disaster recovery environment.

Integration Medium: This hybrid cloud model is based on virtual machine based integration and needs Virtual Private Network (VPN) kind of infrastructure support for moving the virtual machine images quickly between private and public cloud. At times the required data is also required to be moved or synchronized in the form of virtual machine servers, which addresses the data synchronization across the hybrid cloud model.

Challenges: To set-up such hybrid cloud models, it requires additional network topology changes including set-up of VPN, firewall rules, network connectivity and security

Guidance: Since this model is completely dependent upon the virtual machine image movements and extensions, it is important to have the virtual machine image portability across the public and private cloud with less or no format conversion. Since this model would require managing virtual machines across the private and public cloud, it is important to have right management tools. Single pane views, VM management tools deployed on-premise which work across private and public cloud are suggested.

Hybrid Applications & Services

This hybrid cloud model works at application level where the private and public cloud applications interact with each other using integration service layer.

Work load Selection: This hybrid cloud model is suitable when the applications need on-demand scalability and integration. Few examples of scale out using hybrid call could be monthly report calculations or convergence of scientific algorithms. For integration between applications, there are use cases such as mobile apps or APIs running in public cloud which exposes data or services from on-premise applications, data aggregation dashboard, pub-sub based integration between vendors, customers and partners etc.

Integration Medium: The integration medium required for Hybrid Applications & Services are middleware based integration services layer -similar to Enterprise Service Bus or Internet Service Bus running in public cloud. The middle layer provides service abstraction, messaging, publish and subscribe models and accessed from on-premise applications through https based REST or API services.

Challenges: To set-up such hybrid cloud models, it demands Service integration layer overheads to be able to create or expose services from on-premise applications services or data, to be made accessible to public cloud applications. Security over the service bus in terms of authentication, authorization and secured transport are another challenges to be addressed.

Guidance: Since this model works at application integration level it is suggested to consume the integration middle layer from public cloud rather than deploying it on-premise. This would mostly be message and event based calls initiated from on-premise applications which would avoid security issues of external applications calling on-premise services directly. The communication with integration services running on public cloud should be secured using https. Service accounts can be used for authentication and authorization by mapping the service accounts to the application services.

 

Hybrid Data

This hybrid cloud model works at data levels where the public cloud is used as extended or aggregated storage media for on-premise data.

Work load Selection: This hybrid cloud model is suitable when the on-premise data management is expensive due to the size of the data. Data such as archives, backups and rarely used data can be stored in public cloud and used on-demand when required from on-premise applications. Also, data aggregation for business intelligence, analytics, reporting and dashboard can be done through this pattern where the on-premise data is uploaded

Integration Medium: The integration medium is mostly tools oriented which are responsible for identifying and uploading the data to the public cloud. The tools communicate with public cloud storage over secured FTP / HTTP channels with appropriate authentication mechanism. These tools run on the on-premises.

Challenges: The most important challenge in this model is about data security and compliance.

Guidance: Only the public cloud compliant data should be moved in this pattern. It is very important to first classify the data compliance need and accordingly this model is used along with the appropriate public cloud location. Transport security such as SFTP / HTTPS should be used while uploading / downloading the data from public cloud. Suitable data at rest security mechanism such as encryption should be implemented while data is stored on public cloud.

The author is principal architect, Symphony Teleca