Überblick
Moving business-critical infrastructure, software platforms, and applications to the cloud is one of the highest-stakes projects the typical enterprise will ever pursue. Cloud migration is virtually a prerequisite for any form of digital transformation — a reality reflected in cloud adoption’s profound ongoing effects on IT budgets.
In 2020, two-thirds of CIOs ranked public cloud migration and private cloud expansion as the top drivers of their IT spending: a 20% uptick from 2019, per Deloitte. For this spending to ultimately pay dividends, an organization must have a cloud migration strategy backed by solutions that minimize the risks of costly, time-consuming technical complexity and subpar workload performance.
What is cloud migration?
Cloud migration is the movement of workloads — such as an application, a set of physical or virtual servers, or a database within an analytics ecosystem — from an on-premises environment like a data center to the cloud. A migration strategy may also involve movement from one cloud to another.
For analytics workloads in particular, migrating to cloud computing offers unique high-level advantages over sticking exclusively to on-premises infrastructure. The environments from which these workloads are being migrated typically feature legacy infrastructure: hardware and software that are costly to maintain and incapable of delivering — on their own — the speed and scalability that real-time and AI analytics require.
The overall migration process most often targets a public cloud environment, offered and maintained by a major cloud service provider such as AWS, Google Cloud, or Microsoft Azure. If an enterprise is consuming many cloud computing services simultaneously across more than one cloud, then it is pursuing a multi-cloud approach.
Migration can also involve a hybrid cloud, the combination of public cloud and on-premises assets. This setup preserves existing infrastructure and may provide tighter control over performance and cost, while harnessing the power of cutting-edge public cloud services. Data analytics platforms may be migrated to a hybrid cloud environment to maintain such flexibility.
Virtually anything currently running on-premises can be migrated to a public, hybrid, or multi-cloud environment. Business applications and storage are among the most commonly moved workloads. But as public cloud services have evolved, the range of possibilities has expanded to also include numerous artificial intelligence (AI) and machine learning engines (ML), cloud databases, network connectivity tools, and other systems that can support previously on-premises analytics workloads.
What are the different types of cloud migration?
There are multiple ways to migrate workloads to the cloud, ranging from the simple to the complex. The cost and effort required for a successful migration will hinge on the cloud migration tool's ease of use and the different processes for managing data migration. Ideally, a tightly structured multi-phase approach with deep automation will be used to ensure a smooth transition to cloud infrastructure.
The three major migration strategies are:
1. Rehost (aka lift-and-shift)
The simplest option, lift-and-shift entails moving an existing system as-is to a cloud environment, with the old application architecture running on infrastructure-as-a-service (IaaS). Rehosting is relatively inexpensive upfront, but depending on the tools used, it may cost more than other options in the long run. This is due to the lack of optimization for cloud-native APIs and features that can incur opportunity costs as well as excessive operating expenditures on unoptimized workloads. Organizations that succeed with cloud migration generally either:
- Find a platform that is virtually the same regardless of the environment (on-premises, public cloud, multi- or hybrid cloud), making lift-and-shift as efficient as it is expedient.
- Go beyond lift-and-shift to create a new cloud-forward architecture.
Medibank took the rehosting route when it moved over 16 TB of data into Teradata Vantage in the public cloud. Looking to build market-leading health service offerings and experience, the company moved seven applications from providers like SAS, Tableau, and MicroStrategy from its on-premises data warehouse to the AWS cloud, where Medibank now runs them. Because Vantage is the same software in every environment, Medibank did not have to recode anything along the way.
2. Replatform
A slightly more involved approach than rehosting, replatforming entails making some modest changes to a workload before undertaking its cloud migration. Changes in middleware and the use of platform-as-a-service (PaaS) from a cloud provider are both common steps in the replatform approach.
Replatforming may also alter how an application interacts with a database, or how it scales. The introduction of infrastructure-as-code, which allows for resources to be configured and provisioning through machine-readable files like Ansible playbooks, is also possible when replatforming.
As with rehosting, replatforming keeps the core architecture from on-premises intact. It is a somewhat conservative cloud migration strategy, but one that may harness more of the power of cloud computing services than a standard lift-and-shift.
3. Refactor
This approach requires much more time and effort on the part of the organization than the two listed above. Instead of keeping the application more or less the same as it had been on-premises, refactoring requires changing its overall architecture to take advantage of the new cloud environment.
Before refactoring an application or other workload prior to a cloud migration, an enterprise should weigh the costs against the benefits. Refactoring is labor-intensive and costly. However, it can pay dividends in the form of workloads better optimized for the technical capabilities and economic models of the cloud. The approach may also be beneficial for creating a modern data architecture.
A common cloud trend in refactoring is to change a monolithic architecture into a microservices-based one. Microservices make applications easier to modify. Rather than pushing a change that requires retesting the entire interconnected system, teams only need to update the independently maintained and deployed components in question.
The microservices approach also aligns with the emerging concept of data mesh, a data modeling approach organized around business-specific domains, each with its own cross-functional team overseeing how information is managed. Data mesh is capable of enabling increased DevOps agility for cloud workloads.
Other migration possibilities
During the migration process, it may also become apparent that the workload being assessed is either not suited to the cloud or no longer needed at all. These discoveries are sometimes categorized as their own forms of cloud migration, under names like “replace,” “retire,” or “retain.”
For example, a legacy application might be replaced with a software-as-a-service (SaaS) equivalent. Conversely, an on-premises application could be left in place and connected to a hybrid cloud environment.
The 5 biggest benefits of cloud migration
No matter the type of migration, the goals are generally the same: To modernize the workload being moved so that it achieves some combination of better performance, cost optimization, and overall flexibility. The main benefits of cloud migration are:
1. Scalability and elasticity
Scaling a workload on-premises requires purchasing, provisioning, maintaining, and adding or subtracting physical servers on a regular basis. This process is complicated and can quickly become expensive. By contrast, cloud environments provide on-demand computing that can be scaled up and down as requirements change. Elasticity goes a step further by scaling dynamically to match cloud resources to rapidly evolving workloads.
Tip for analytics: Use a single analytics environment that offers scalable data exploration, modeling, and scoring, on cloud infrastructure of choice and with hybrid deployment support.
2. Cost reduction
Cloud adoption affects costs in multiple ways. Capital expenditures on hardware and software turn into operating expenditures on cloud service subscriptions. Fixed costs become variable. Yet the pay-as-you-go nature of cloud computing can, at times, be a double-edged sword. Unpredictable shifts in workloads can send costs soaring. Moreover, without software-based optimization, workloads won't strike a balance between performance and cost.
Tip for analytics: Look for analytics solutions with advanced indexing techniques, which provide savings versus brute-force alternatives that rely entirely on hardware. Also, seek out a blend of predictable and pay-as-you-go pricing options so money isn't wasted on unused capacity.
3. Flexibility
Migrating to the public cloud doesn't have to mean leaving on-premises behind. Cloud deployment options are highly flexible and include architectures such as hybrid cloud and hybrid multi-cloud. Existing investments can be maximized even as a company taps into the virtually limitless resources available through a public cloud. The cloud as a whole also simplifies access to critical resources, including advanced AI and ML tools.
Tip for analytics: Use software that works similarly on-premises and in the cloud. This simplifies migration and helps uncomplicate the operation of hybrid deployments. Don't settle for a cloud-only data warehouse that limits possibilities out of the gate.
4. Performance
Moving from on-premises to the cloud can be a performance upgrade, if the previous environment was noticeably aging and/or physically far from its users. The cloud also delivers new capabilities, like autoscaling, that help ensure acceptable performance as demand ramps up. Hybrid clouds can preserve the local performance of on-premises hardware and software when necessary.
Tip for analytics: Utilize a platform with dynamic resource allocation to consistently meet the service level agreements of both strategic and tactical workloads. Separate scaling of compute and storage capabilities is also critical for providing appropriate resources at an affordable cost.
5. Expanded functionality
The major cloud service providers are on the cutting edge of every domain, from AI and ML to cybersecurity. Accordingly, cloud customers have access to an ever-expanding set of features, the breadth and depth of which they couldn't attain if operating exclusively on-premises. Numerous built-in services simplify the path to workload modernization and optimization.
Tip for analytics: Seek integrated AI and ML in a cloud data analytics platform. Otherwise, there is the complexity of connecting a third-party partner engine for the same purposes.
5 common challenges in cloud migration
It is common for a cloud migration strategy to run into roadblocks along the way, particularly if an organization hasn't chosen a data analytics platform optimized for today's connected multi-cloud reality. Some recurring challenges include:
1. Unchecked costs
Although a cloud migration should save money on paper, it may actually increase costs in some cases. For example, cloud-only data warehouses have limited performance optimizations, meaning that the only way to deal with many performance issues is to simply buy more clusters, causing costs to balloon rapidly.
2. Application refactoring
Retooling an app for a cloud environment can improve its performance and make it more cost-effective, but significant effort is required to do so. The necessary time investment can have ripple effects on a company's operations, as it may struggle to stay on schedule with the migration and to properly control downtime.
3. Database migration
Accenture has highlighted the complexities that database movement entails, including the care required to transition these enormous amounts of data, the complexity of the legacy environment, and the need to modernize throughout the course of the migration. Plus, after the migration finishes, there is still the challenge of making sure that all the data has retained its integrity.
4. Partner selection and management
There are many cloud vendors on the market, and it's critical to pick one with the right range of services for an enterprise's specific business needs. Without proper vetting of cloud partners, it's possible to end up locked into a particular provider's services, or you could be forced to piece together numerous point solutions that aren't well-integrated.
5. Planning
For companies sitting down to map out the cloud journey, the task can feel daunting. Above all, the goal of migration is to modernize your enterprise, but many other questions must be carefully considered when planning how to achieve that goal. How will the organization maintain ample uptime during the migration? What will it do to encourage user adoption of new cloud-based tools? How can costs be controlled and budgets set? A migration must be carefully planned at the outset — it's not meant for an improvised approach.
Making cloud migration work
Successful migration strategies require the right combination of people, processes, and technologies within a cloud migration service:
- People: Consultants can provide in-depth guidance and build custom solutions matching the needs of each enterprise.
- Processes: A multi-phase approach, beginning with inventory and baseline assessments and then proceeding through business process modeling to recommendations.
- Technologies: A connected multi-cloud platform enables organizations to use their preferred cloud(s) and technology stacks for analytics.