As Federal technology managers increasingly operate in a hybrid, multi-cloud world, having a holistic view of their agency’s application landscape and architecture will help in mapping the right workload to the right type of cloud deployment.
Tech managers must understand the demand pattern of an application as well as key metrics–such as cost, CPU performance, connectivity, security, storage, and scalability requirements–to determine which cloud hosting model is most cost-effective.
As technology managers migrate workloads to cloud infrastructures, they should “consider the attributes or characteristics of the different environments and what types of workloads are suited to [run in] the environment,” according to Melanie Posey, research vice president for 451 Research’s Voice of the Enterprise: Cloud Transformation offering, during a webinar.
The 451 Research Group offers organizations advice that puts in context the many different cloud deployment choices and the different workloads that might be suitable for those deployment models–on-premise private cloud, hosted private cloud, public infrastructure-as-a-service cloud, or public software-as-a-service environments.
For example, the on-premise private cloud is suited for purpose-built legacy applications, IT governance concerns, data residency requirements, data-intensive or latency-sensitive workloads, and steady-state workloads. Internal IT workloads, Enterprise Resource Planning (ERP) systems, financial management, application development, databases, data warehouses, and industry apps could run in these cloud infrastructures.
Hosted private clouds are suited for situations in which operating costs are preferred over capital outlays, and there is a lack of in-house cloud management skills, steady-state external-facing workloads, and distributed user population. This could encompass e-commerce/customer support, digital media, social applications, and disaster recovery/business continuity workloads.
Public IaaS clouds are geared for workloads that require on-demand and elastic scalability, process-intense computing, highly distributed and external-facing workloads, and/or a cloud-native development approach. This could include big data analytics, virtual desktops, industry-specific solutions, digital media publishing, and mobile apps, basically “peak load” or scale-up types of applications.
Public SaaS clouds are suited for what 451 Research describes as no-ops environments, as well as version control needs, standardized identity access and management, managed scalability and complexity challenges. This includes applications such as email/collaboration, digital media, point-of-sale, ecommerce, customer relationship management, and human resource management.
It’s important to note that not every workload is meant for cloud migration–and these examples might not necessarily fit all applications. That’s why it is important to identify, inventory, and then rank applications by migration potential, according to cloud experts. There are many migration options, including rehosting, refactoring (running an organization’s applications on a cloud provider’s infrastructure), revising, rebuilding, or replacing. Once tech managers decide on an application migration strategy that is conducive for their agency, they must apply predictive analytic tools that will help them gauge how those applications will perform in the cloud based on analysis of present performance and usage. Then they can project the performance of each workload’s compute and storage resources in hypothetical scenarios. Additionally, they must map out all the applications and systems the applications interact with as well as predict the costs of running their applications in the cloud.
Migrating workloads to the cloud requires a step-by-step process in which the demand pattern of an application as well as key metrics such as cost, CPU performance, connectivity, security, storage, and scalability requirements, are critical for success. And it all begins with picking the right kind of cloud deployment to match the required workload.