The fundamental value resulting from data-driven processes has become increasingly linked with analytics. Once considered a complement to intuitive decision-making, analytics has emerged as a main focus of mission-critical applications across industries for any number of use cases.
Yet, as the purposes for utilising analytics for business processes have increased, so has the complexity of deployments. Organisations must now habitually confront circumstances in which data is spread across a plenitude of environments, making it arduous, error-prone and time-consuming to try to centralise for a single use case. Perhaps even more widespread is the reality in which it’s beneficial to deploy in multiple settings (such as with Linux platforms, in the cloud, or with containers), but budgetary or technological shortcomings make it unviable. Certainly, application performance oftentimes suffers as well.
The fact is, we live in a world business climate that is continuously evolving, which consequently continuously evolves our data space – and this calls for enterprise agility for analytics as much as for any other aspect of competitive advantage. Processing is optimised by performing analytics as close to data as possible, which may need to change locations for scheduled downtime, disaster recovery (DR), or other business requirements such as seasonal or limited-time offers in the cloud.
By embracing an agile approach, built on a foundation of what can be called “intelligent availability,” organisations can dynamically provision analytics in a virtually limitless list of heterogeneous environments to satisfy countless business use cases, seamlessly and rapidly moving data between on-premises settings (including both Windows and Linux machines), the cloud and containers.
Accordingly, organisations benefit from decreased infrastructure costs, effective DR, and an overall greater yield for analytics—and that of data in general.
Analytics in the cloud
One of the more widespread approaches in which intelligent availability improves analytics is with cloud deployments. There are numerous benefits to moving to the cloud for analytics, not the least of which are the pay-per-use pricing model, less infrastructure, and elastic scalability of cloud resources. There are also several software as a service (SaaS) and platform as a service (PaaS) options—some of which involve advanced analytics capabilities for machine learning and neural networks—for users without data science experts on staff. Nonetheless, the most persuasive reason for running analytics in the cloud is facing the alternative: attempting to scale on premises. Customarily, scaling in physical environments involved an exponential curve with numerous unalterable costs which frequently limited application performance and enterprise agility. By scaling in the cloud and with other contemporary measures, however, organisations enjoy a far more affordable linear curve.
This point is clearly demonstrated by a healthcare example in which a well-known, global healthcare organisation was using SQL Server on premises for its OLTP, yet wanted to deploy a cloud model for Business Intelligence (BI). The choice was clear: either ignore budget constraints by indulging in additional physical infrastructure (with all the unavoidable costs for licenses and servers) or deploy to the cloud for real-time data access of their present kit. The latter option decreased costs and maximised operational efficiency, as will the majority of well implemented analytics solutions in the cloud.
In this example and countless others, optimising cloud analytics involves continually replicating on-premises data to the cloud. Smart organisations diminish these costs by choosing asynchronous replication; the previously mentioned healthcare institution did so with about a second latency for near real-time access of its healthcare data. Replication to the cloud is often inexpensive or even free, making the data transfer component highly cost-effective. By making this data available for BI in the cloud, this organisation effected several advantages. The most prominent was the reproducibility of a single dataset for multiple uses. Business users—in this case physicians, clinicians, nurses, back-office staff, etc.—are able to access this read-only data for intelligence to impact diagnosis or treatment options, as well as for administrative/operational requirements (OLTP).
This latter point is very important. With this paradigm, there are no application performance issues compromising the work of those using on-premises data because of reporting—which could occur if each group was provisioning the same copy of the data for their respective uses. Instead, each user benefits mutually from this model. The healthcare group is assisted by the primary data being stored on premises, which is important for compliance measures in this highly regulated industry. It’s also important to note the flexibility of this architecture, which most immediately affects cloud users. Organisations can establish clusters in any of the major cloud providers such as Amazon Web Services (AWS), Azure, or any private or hybrid clouds they like. They can also readily transition resources between these providers as they see fit: feasibly according to use case or for discounted pricing. Even better, when they no longer need those analytics they can speedily and painlessly halt those deployments—or simply migrate them to other environments involving containers, for example.
Plus automatic failovers
The above-mentioned healthcare group also gets a third advantage when utilising an intelligent availability approach for running analytics in the cloud: automatic failover. In the event of any sort of downtime for on-premises infrastructure (which could include scheduled maintenance or any sort of catastrophic event), its data will automatically failover to the cloud using intelligent availability techniques. The ensuing continuity enables both groups of users to continue accessing data so that there is no downtime. Those primary workloads simply transfer to cloud servers, so workloads are still running. This benefit typifies the agility of an intelligent availability approach. Workloads are able to run continuously despite downtime situations. What’s more, they run where users specify them to create the most meaningful competitive advantage. Most high availability methods don’t provide users with the flexibility of choosing between Windows or Linux settings. There’s also a simplicity of management and resiliency for Availability Groups facilitated by intelligent availability solutions, which provision resources where they’re needed without downtime.
Intelligent availability technologies and methodologies allow users to maximise analytic output by generating recurrent advantages from what is fundamentally the same dataset. They allow users to migrate copies of that data to and between cloud providers for low latency analytics capabilities, with some of the most advanced techniques available today. What’s more, this approach does so while maintaining critical governance and performance requirements for on-premises deployments. Perhaps best of all, it maintains these benefits while automatically failing over to offsite locations to preserve the continuity of workflows in an era in which information technology is anything but predictable.
Don Boxley, CEO and co-founder, DH2i
Image Credit: Sergey Nivens / Shutterstock