Hybrid cloud architectures play a vital role in helping enterprises migrate safely to the cloud and provide a way to meet data governance and risk management requirements for data that must stay on premises. But moving applications and the databases that support them to even a hybrid cloud requires considerable planning and testing, plus ongoing management and monitoring.
In a hybrid cloud database environment, some data is stored and managed on premises and some is moved to the public cloud. As a result, running databases in a hybrid cloud introduces new data management considerations that must be addressed to keep data secure, accurate and compliant with regulations, while also ensuring it can be processed efficiently.
What is a hybrid cloud?
A hybrid cloud combines public cloud services with traditional enterprise IT infrastructure that's typically set up as a private cloud. The enterprise IT aspect could include servers managed directly by an organization in its own facilities or colocated in a third-party data center that's shared with other users. Some applications are managed using familiar in-house IT processes, while others are managed through cloud-specific processes.
Hybrid cloud database deployments extend the same conception to the data itself. But the decision to go to a hybrid cloud database model should be driven by application and workload needs that call for databases to be on a hybrid cloud, said Yugal Joshi, partner at technology research advisory Everest Group. If that's the case, hybrid architectures can provide complementary benefits to the underlying applications by streamlining access to required data.
Although they often provide lower costs and more flexibility than on-premises database systems, cloud services aren't practical for every enterprise or application. "With increased data scrutiny, strong data gravity, workload requirements on latency, licensing complexity and data decentralization, not every data can be put on one source such as public cloud," Joshi said. "This is where a hybrid model can add value."
Benefits of a hybrid cloud for deploying databases
The benefits of hybrid cloud database environments are like those of hybrid clouds for applications: providing access to automated cloud services, opening new options and improving portability.
Automated cloud services. "By using a hybrid cloud to deploy databases," explained Alexander Wurm, research analyst at advisory Nucleus Research, "organizations can reap the benefits of the modern cloud, such as regular updates and elastic scalability, without interfering with the security and reliability provided by existing on-premises infrastructure in support of mission-critical workloads."
New options. Enterprises can also explore new options that were not practical with legacy database technology. If needs related to key variables like security, performance, quality or cost change over time, more options are available to rebalance the portfolio, said Joshua Swartz, a partner in the digital transformation practice at management consultancy Kearney.
Portability. A hybrid cloud database approach also supports data and workload portability across multiple private and public cloud services. "This, in turn, allows an organization to pick and choose coordination of data and workloads across hybrid cloud, thereby avoiding vendor lock-in, achieving cost and efficiency benefits, and the ability to grow and shrink environments to meet service demands," reasoned Terri Sage, CTO at data management and analytics platform provider 1010data. In addition, portability can make recovery and business continuity planning easier and encourage experimentation and innovation.
What to consider when planning a hybrid cloud database strategy
Hybrid cloud architectures provide benefits that can unlock new opportunities to lower costs; however, they also introduce new security, performance, integration and data quality challenges that need to be initially addressed to maximize a hybrid cloud database strategy. IT teams, data managers and database administrators should weigh the following issues before deploying databases in a hybrid cloud environment.
1. Digital transformation and application modernization goals
One of the best starting points is to flesh out the various goals to modernize and transform business processes and the applications that power them. "Organizations need to understand the digital transformation goals for not only the business, but also the outcomes they want from modernizing the existing applications and the databases used by those applications," said Brian Schneider, managing delivery architect for multi-cloud at IT management consultancy Capgemini. The result should be the most efficient and effective database option for the business and end users.
This process should begin with a discovery phase that includes application teams and business owners to determine the current architecture, application experience and end-user pain points, followed by the creation of a transformation roadmap for making improvements. Getting stakeholders to participate in the process is essential. Data managers can help stakeholders stay abreast of on-premises and cloud database technology advancements that might affect planning.
2. Appropriate grouping of applications and databases
Focusing on business and application goals can also help identify the best way to stage data to support different application requirements. "The movement of applications and databases requires the appropriate grouping of the applications and databases into logical units," noted Colin Dawes, CTO at managed service provider Syntax.
Creating these natural fault lines can help data management teams divide monolithic systems into manageable chunks. Getting this part of the process wrong, Dawes warned, can present performance and stability issues that result in an overall rejection of the process by the stakeholders.
3. Cost-benefit analysis vs. other approaches
Data managers need to analyze the relative costs and benefits of modernizing their existing on-premises database, migrating to the cloud or pursuing a hybrid approach. A hybrid cloud will be inherently more complex and expensive to deploy and manage than a cloud-only or on-premises approach. "The additional costs and management overhead add up and need to be justified by benefits and business requirements," said George Chedzhemov, senior vice president of client services at data intelligence platform provider BigID.
The added expense may be worth it, but enterprises also need to carefully weigh the challenges of enlisting new cloud database services that create additional headaches as part of a hybrid strategy. Chedzhemov argued that proprietary approaches, such as AWS DynamoDB or Google Cloud Spanner, can limit deployment options. He recommended cloud services based on open standards such as MySQL, PostgreSQL, MongoDB and Apache Cassandra to improve compatibility across on-premises and cloud services.
4. Data egress charges
A hybrid cloud database strategy should include the flow of data. Data transfer costs that were negligible with an on-premises database could be high with the move to a hybrid environment. "These costs," Sage noted, "can be significant and depend on the amount of data replicated in support of a hybrid cloud database strategy."
Some of these costs can be eased with the proper architecture. Nonetheless, appropriate controls should be implemented if the data flows over expensive channels.
5. Data latency
A hybrid cloud database can also introduce network latency due to data transfer between the different cloud service providers and the distance between physical resources. A hybrid approach typically results in longer routes and more network hops, which can add several milliseconds or even seconds to data transfers, Sage said. Planning should include accounting for network latency and revisiting decisions on the physical regions selected. "Sometimes it will make more sense to colocate different cloud service providers in similar geographical areas to reduce both cost and network latency," she advised.
Configuration choices also need to be considered from a latency perspective when planning whether a cloud or on-premises node has a more active or passive role. Active-active cluster configurations, for example, often have less competitive latency between private and public clouds, Wurm said, adding that active-passive configurations may be a better choice for asset-intensive industries with an abundance of edge data.
6. Data security
Governing and protecting data must be part of a hybrid cloud database strategy since the complexity of hybrid cloud environments can increase potential attack surfaces, said Brian Platz, CEO and co-founder of blockchain database platform provider Fluree. "[I]t is essential," he explained, "to map out the architectural flow of data across all possible environments and put into place security and governance measures that account for data as it is managed, deployed, ported and virtualized across all possible scenarios."
Consider using continuous integration/continuous delivery testing and version control to mitigate security risks. It may also be worthwhile to explore data-centric security governance that protects the data as it moves across various networks and clouds.
7. New tools and skills requirements
Hybrid cloud databases may introduce new data workflows that need to be addressed. Data management tool sets might vary across public and on-premises systems, which can increase the cost of operation, Everest Group's Joshi said. He suggested developing a standard operating model and tool strategy for scaling, cross-skilling and plug-and-play operations.
Along those lines, different skills may be required to support these new workflows. "Finding talent for public cloud is tough," Joshi acknowledged, "but it is even worse for hybrid cloud."
8. Balancing stability and simplicity
Any new infrastructure for storing and transferring data has the potential to create a new point of failure. Consider how to minimize disruptions to operations when a system or network goes offline. "Solving this is a bit like an insurance policy," Kearney's Swartz said. "One can absolutely create redundancy and fail-safe mechanisms, but the costs are quite high." The approach most companies follow is to tier data based on business criticality and provide the costliest redundancy for only the most critical data.
It's also important to provision for the work needed to integrate multiple systems. Every additional system or database brings another interface necessary for integration with core applications and systems. Developing an architecture with fewer interfaces can lower management risk.
Managing a hybrid cloud environment can be far more complex and drive higher costs, compared to a cloud-only or on-premises approach, Swartz said. Changes, updates, patches and enhancements all require more extensive and elaborate planning, testing and monitoring to avoid creating a domino effect of compatibility issues.