Data has become one of the most valuable assets in the enterprise. IT teams must make changes -- both culturally and technically -- to ensure their strategy reflects that.
As the world generates more data than ever, it strains the infrastructure needed to process, transfer and store it. Traditional data management approaches can no longer keep up with demand, and organizations must find other ways to accommodate increasingly complex and data-intensive workloads.
The data management industry has been bombarded by change. Not only do organizations generate more data, but that data varies in format and is more widely distributed. Enterprises also contend with an influx of data-related regulations, while they address consumer demands for quicker and more complete access to data. At the same time, businesses try to derive greater value from all this data, as reflected by trends such as big data and advanced analytics.
Data plays an essential role in just about every aspect of day-to-day life, whether it be healthcare, shopping, entertainment or home utilities. Under such circumstances, organizations that handle all this information must embrace an architecture that reflects the pivotal role data now plays in application and service delivery.
Ditch the application-centric paradigm
Until recently, most organizations have taken an application-centric approach to data management and infrastructure planning. Their primary concern was to build applications that provide users with the services they need. They often treated data as an afterthought. From this perspective, data exists only to support the applications. Each application adheres to its own data model and comes with its own data store, whether a relational database, key-value repository or other storage format.
The application-centric approach typically results in redundant data and duplicated efforts. It also makes it time-consuming and expensive to upgrade applications or implement new ones. When an organization plans an application, it often gives little thought to how it might reuse or derive other benefits from the data. The application maintains complete control over the data, with the data itself treated as a second-class citizen. A single organization might support hundreds or even thousands of applications.
Organizations that adhere to the application-centric model typically contend with data silos, which can introduce data inconsistencies and errors. In such an environment, security teams struggle to safeguard data and ensure compliance. In addition, departments and individual teams often have their own segmented views of the data. Stakeholders can find it difficult to access the information they need or to perform comprehensive analytics. Companywide strategies can be complicated, time-consuming and expensive to implement, especially when trying to share, cleanse or transform the data.
To compensate for the limitations of the application-centric approach, many enterprises have implemented, or have tried to implement, data warehouses, data lakes or other big data-related technologies and projects. However, these approaches don't address the underlying challenges that come with an application-centric architecture. In addition, such projects can be difficult and expensive to deploy and maintain. In fact, many of them have been largely unsuccessful, with some being scrapped before ever hitting production.
What is a data-centric architecture?
Organizations are moving toward a data-centric model because of the limitations of the application-centric approach. The new paradigm might be referred to as data-centric architecture, data-centric infrastructure or data-centricity, but the goal is the same: move data to the front of the line where it belongs.
Many online companies have taken a data-centric approach since their inception. They organize their infrastructure and services in a way that maintains the data structure, despite how individual applications are deployed, updated, replaced or retired. In this model, applications are transitory, but the data endures.
A data-centric approach is as much a cultural shift and mindset change as it is a physical implementation.
With the data-centric approach, IT teams design infrastructure around data requirements, rather than trying to make data fit the infrastructure. Although cost and operational ease are still important, the data becomes the primary consideration in determining how to deploy infrastructure. However, a data-centric architecture affects more than just infrastructure planning; it can also affect application development and business decisions. For example, when an organization adopts a data-centric strategy, it builds applications according to the prevailing data model, rather than developing a specific model for each application.
A data-centric approach is as much a cultural shift and mindset change as it is a physical implementation. Rather than put applications above all else, look for ways to optimize data management and delivery strategies. Then, build applications to be consistent with those strategies. Not only does the data-centric approach eliminate silos, it simplifies data management and application development because everyone works within a common data framework.
Become a data-centric organization
To adopt a data-centric architecture, treat data as a central asset that will be around long after applications and infrastructure. Strive to eliminate siloes, multiple instances of data and the complex sprawl that goes with an application-centric strategy. Implement a single data model for multiple applications that is extensible enough to accommodate specific requirements. This approach doesn't necessarily require a single database or data store, but rather a common vision of the data that offers a unified definition.
Data-centric infrastructure makes it easy to share and move data. It supports federated queries across multiple repositories. It delivers high availability and reliability. It ensures the organization can access the data on demand and in real time. In addition, a data-centric infrastructure enables organizations to share and deliver their data when and where it's needed. It provides the necessary performance, capacity, scalability and security. The infrastructure can evolve to meet changing workload requirements and adopt new technologies.
In the past few years, vendors introduced products that promise to help organizations achieve their data-centric goals. Composable disaggregated infrastructure is flexible enough to meet evolving data requirements. For example, in 2015, Hewlett Packard Enterprise introduced its Synergy composable infrastructure appliance, which deploys IT resources quickly and automatically to accommodate changing demands.
Individual devices can also provide the necessary components to meet data-centric challenges. Intel, for example, pitches its Xeon processors and Optane memory and storage devices as technologies to accelerate "innovation in a data-centric world." These products, according to Intel, enable users to store, process and move data across demanding workloads, including those at the edge and in multi-cloud environments. In other words, the devices deploy infrastructure according to data-driven requirements.
Fungible is another company to adopt the data-centric vision. The vendor offers a line of data processing units (DPUs) intended to address two issues with scale-out, disaggregated data centers: the inefficient execution of data-centric computations within server nodes, and the inefficient interchange of data among nodes. To this end, Fungible has incorporated its DPU into the Fungible Storage Cluster -- all-flash NVMe-oF disaggregated storage -- that, according to the company, can achieve speeds approaching NVMe direct-attached storage, while it processes data-centric workloads more efficiently.
Devices alone, however, do not make a data-centric infrastructure. First, put data at the core of an IT strategy and, from there, determine how to implement the necessary infrastructure.
