As enterprises accelerate their digital transformation, they collect large volumes of data about their business processes. These data, when properly analysed, enable unprecedented improvements in business workflows, along with opportunities for educated and more efficient decisions. Therefore, a great deal of the added value of enterprises’ digital transformation stems from data analytics. As business datasets grow in volume and diversification, enterprises consider cloud-based solutions for executing their data analytics functions. Cloud analytics come with several compelling value propositions, which is the reason why many enterprises choose to ride the wave of cloud computing for their analytics. Nevertheless, the deployment of cloud analytics solutions is associated with various technical and organizational challenges that enterprises must consider prior to investing in data analytics workflows in the cloud.
Leveraging the Benefits of the Cloud
The main benefits of cloud analytics stem from the nature of cloud computing. Specifically, they include:
- Scalability: Cloud computing enables scalable analytics deployments that can accommodate the ever-growing data volumes. With a cloud deployment at hand, enterprises need not worry about the availability of storage and computing resources: The appropriate and adequate amount of resources will be there when needed.
- Elasticity: Cloud resources are elastic i.e., end-users can get more when needed e.g., to accommodate seasonal spikes in the data that must be analysed. At the same time, they can also use less resources when less are needed.
- Pay-per-Use: In a cloud computing context, companies pay only for what they use. This provides flexibility in the cost structure of cloud analytics applications. In this way companies need not undertake risky capital investments in computing infrastructures. Rather they can pay for what they use as they grow and see their revenues increasing.
- Analytics tools availability: Nowadays cloud providers offer enterprises access to a rich set of cloud analytics tools, ranging from business intelligence and reporting tools to machine learning environments. Thus enterprises joining the cloud are provided with many opportunities for analysing data in different ways without any need to install, deploy, and integrate analytics tools.
In addition to these benefits, modern cloud infrastructures come with a host of tools that facilitate application development. These tools offer high-level and easy-to-use functionalities, which decouple IT programmers and business users from the low-level deployment details of the analytical tools. This is nowadays possible because cloud providers offer services at higher levels of abstraction.
Rising Levels of Abstraction
Once upon a time, cloud computing providers supported analytics applications based on elastic access to computing and storage resources based on the popular Infrastructure-as-a-Service (IaaS) paradigm. With IaaS, companies gained access to computing cycles and other hardware resources needed for implementing analytics functions. Furthermore, since the early days of the cloud, cloud providers offer access to Virtual Machines (VMs) in the cloud, which provides abstraction of the operating system. Using VMs, companies can deploy their cloud analytics over the virtualized infrastructure of the providers.
Over the years, the above-listed models evolved in ways that facilitate the tasks of analytics developers and data scientists. Specifically, it is currently possible to run entire application containers in the cloud, which combine entire operating systems, analytics tools and analytics applications in a single package. Such packages decouple developers and deployers from the need to deal with the low-level operations of the platforms (e.g., operating systems and tools installations) that are integrated in the package. Likewise, they facilitate the distribution of the cloud analytics applications using container images. Recently, cloud providers are also offering integrated analytics environments over their infrastructures. For instance, it is possible to use complete machine learning and data analytics environments in the cloud, to develop and deploy analytics applications. This model is characterized as Machine Learning as a Service (MLaaS) since it enables the creation and deployment of entire machine learning pipelines over cloud resources. Along with MLaaS, there are also other Platform as a Service (PaaS) paradigms for cloud analytics, which enable developers to implement end-to-end cloud analytics workflows that orchestrate multiple cloud analytics functions.
During the last couple of years, higher level abstractions have also emerged: Companies are able to invoke pre-trained and deployed cloud analytics functions as serverless cloud programs such as Cloud functions. This paradigm is conveniently called Function-as-a-Service (FaaS). FaaS enables the execution of cloud analytics as serverless functions. These high-level cloud analytics abstractions enable enterprises to save efforts and costs when opting for cloud analytics applications instead of developing their own ones on-premise.
Addressing the Migration Challenges
Despite the benefits of cloud computing, several companies have second thoughts about migrating to the cloud for their analytics. The main reason for this is that most companies face challenges when realizing the migration from on-premise analytics to cloud analytics. Some of the most prominent challenges are:
- Loss of Control: Enterprises are deeply concerned about losing control of their data analysis functions. The adoption of cloud analytics implies that some functions will be provided by the cloud provider and will not be under the control of the enterprise.
- Security and Trust: There are still security and trust barriers when considering cloud adoption. Some enterprises are not willing to move their corporate data to a virtual data centre of a provider, due to the fear of having them managed by a third-party. In most cases, cloud providers offer superior data security, when compared to on-premises security services provided by non-IT companies. Nevertheless, there are still certain types of organizations (e.g., banks and critical infrastructure operators) that prefer to keep their data and analytics in-house.
- Less Innovation Opportunities: When using a cloud provider’s tools, you inevitably become bounded by their capabilities. This provides fewer opportunities for innovating in directions that are not easily supported by off-the-shelf data analytics environments. In an era where cloud analytics are not only about reporting KPIs, but affect the development and deployment of innovative workflows, enterprises need to overcome the danger of becoming framed to the capabilities of a single toolset.
- Need for Proper Skills: Cloud analytics require employees with specific skills, such as knowledge on cloud computing, networks, and cloud models like PaaS, FaaS and Software as a Service (SaaS). To this end, many enterprises invest on upskilling and reskilling their employees, prior to starting a cloud analytics adoption journey. In several cases, this upskilling is part of a cultural shift that organizations need to realize in order to benefit from cloud analytics applications.
- Smooth Migration Paths: Most enterprises cannot realize a transition from on-premises analytics to cloud analytics overnight. Migration is usually a tedious and time-consuming process that has to deal with many different aspects, including data migration, mapping of on-premise developments to the capabilities of cloud tools, as well as the implementation of on-premise data processing workflows as cloud-based pipelines. To alleviate these challenges, there is a need for defining smooth migration paths. The latter may involve the establishment of hybrid cloud environments that combine cloud analytics with on-premise analytics, as an intermediate step prior to the full migration.
Overall, Chief Information Officers (CIOs) and Senior IT Managers cannot afford to ignore the power and the benefits of the cloud when planning their business analytics functions. However, they must also make provisions for addressing the technical, technological, and organizational challenges presented above. Migrating to cloud analytics requires effective technology management to deliver the promise of increased efficiency at a lower cost.
