Moving analytics to the cloud
Reap the same benefits for analytics as organizations do for many other applications
Steve Holder, National Strategy Executive Analytics & AI at SAS
It is clear that cloud computing has become mainstream. Cloud has navigated across Gartner’s hype cycle, and has arrived at the Plateau of Productivity. The cloud has become so common place that as the “technology guy” in my family I have aunts and uncles asking how they can move to the cloud. It has become virtually (pun intended) synonymous as the way for organizations to break down IT barriers to deliver the much-needed agility, elasticity and speed.
The advantages of digital transformation and faster time-to-value often are allocated only to cloud-native applications. And, generally, an organization’s analytics applications are not cloud-native. How do we better reap the benefits of cloud computing for our analytics workloads?
That was the focus of a recent installment of our Analytic Executive Quarterly Breakfast Series in Toronto. A few dozen analytics leaders joined us to hear the thoughts of panelists Charles Victor, Senior Director of Strategic Partnerships and External Services at the Institute for Clinical Evaluative Sciences (ICES); Jonathan Carroll, Chief Information Officer of ERPM North America Commercial Banking for the Bank of Montreal; and Gavin Lubbe, a partner with KPMG specializing in large-scale insurance deployments.
The cloud model, on the surface, is tailor-made for analytics. Traditionally, we build out infrastructure to accommodate our heaviest demands. If that demand is a huge analytic workload that runs infrequently, but regularly, having elastic computing resources could help manage costs better.
Are you migrating your analytics to the cloud?
The advantages of digital transformation and faster time-to-value often are allocated only to cloud-native applications. And, generally, an organization’s analytics applications are not cloud-native. How do we better reap the benefits of cloud computing for our analytics workloads?
That was the focus of a recent installment of our Analytic Executive Quarterly Breakfast Series in Toronto. A few dozen analytics leaders joined us to hear the thoughts of panelists Charles Victor, Senior Director of Strategic Partnerships and External Services at the Institute for Clinical Evaluative Sciences (ICES); Jonathan Carroll, Chief Information Officer of ERPM North America Commercial Banking for the Bank of Montreal; and Gavin Lubbe, a partner with KPMG specializing in large-scale insurance deployments.
The cloud model, on the surface, is tailor-made for analytics. Traditionally, we build out infrastructure to accommodate our heaviest demands. If that demand is a huge analytic workload that runs infrequently, but regularly, having elastic computing resources could help manage costs better.
We’ve seen organizations deploy three strategies for moving their analytic workload into the cloud, and each has considerations and constraints given the complexity of analytics applications.
- The “lift-and-shift” approach: Taking the entire computing infrastructure and moving it to a hosted environment retains the inefficiencies of an on-premise environment and doesn’t take advantage of cloud-native features.
- Adoption of open source technologies for analytics: Moving all analytic workload to open source can add flexibility, but introduces complexity in terms of development standards, deployment and governance of your analytic assets.
- Recoding analytics to run in the cloud: Analytics workloads in a cloud-native environment are resource-intensive, sometimes without a clear value proposition. On top of that is the issue of data residency; if your data isn’t in the cloud, it takes a huge pipe to applications running there.
While each of these is viable in isolation, as you layer in the complexity of most organizations data and analytics landscapes the ability to deliver choice, control and scale, you can quickly lose the benefits of cloud.
Enter containerization.
Until the 1930s, shipping docks relied heavily on “swampers.” Swampers were labourers who were employed to jam cargo into ships, fitting as much as possible into the hold. Then came standardized shipping containers. Regardless of what was in them, shipping capacity was more predictable, the process of loading a ship became more orderly, and the shipping industry was changed forever.
The same concept has been applied to cloud computing. Containerization of compute workloads packages up an application and all its dependencies inside a virtualized operating system or container. They’re independent, portable, and run on any private or public cloud platform. While as little as five per cent of enterprise workload is containerized today, we expect adoption of 30 to 40 per cent in the future.
Implementation of a container strategy can help bring cloud attributes to any landscape regardless of where the data and applications reside. While containers mitigate the risks of migration to cloud-based analytics, they aren’t a panacea. There are plenty of issues that demand attention and monitoring.
“We have a close eye on every instance of how we migrate to the cloud,” says Charles Victor from ICES. The organization facilitates collection and use of health data for research purposes. It’s sensitive data with the strictest of privacy standards, and Victor’s job is to enable research partners to access it. Data residency is a pressing concern. That concern is ameliorated by the use of a private cloud shared with Toronto’s Hospital for Sick Children High Performance Computing program. Its infrastructure can hold the entire data repository. It’s expensive, and it’s not in use all the time. And it’s a “unique, but peer” partner held to the same data security standards.
CLOUD MANAGEMENT
Maintaining data integrity can also become challenge when team members can spin up containerized instances of applications and all their dependencies. It’s important to prevent duplication and conflicting versions of data moving from container to container, says BMO’s Jonathan Carroll—the computer-age-old mantra of the single version of the truth. There’s value in sandboxing and experimenting with payloads, but there’s a responsibility to “write back” to reconcile the data. ICES has a dedicated data quality team that works closely modeling and production teams, Victor says.
As in any virtualized environment, resource management must be addressed. One of the major advantages of cloud-based computing is cost management. Compute cycles cost money. If someone spins up an instance for sandboxing, for example, and forgets to spin it down, there can be sticker shock when the bill arrives and budgets are already set.
Deployment of analytics in a cloud environment comes down to processes and people, and in terms of the latter, Canada is uniquely gifted, according to KPMG’s Gavin Lubbe. He has directed mammoth cloud deployments in insurance and financial services; his experience with projects in South Africa, the United Kingdom and Canada gives him an informed view of the skills landscape.
“We’ve got enormously good talent,” he says. “The universities are doing a wonderful job” of churning out people with the required skill sets. The next step is to become more multi-disciplinary—“all can code, all understand the business, all understand problem solving.”
That kind of holistic talent would go a long way to better integrating the modeling team with DevOps, an important consideration, according to all three.
SERVICE STACKS
Cloud computing lives at many points in the enterprise computing stack—organizations can choose how much of the infrastructure they want to outsource to a provider (internal or external), right up to the content layer of the stack. That wholesale service stack provision is what we at SAS call Results-as-a-Service (RaaS). The customer provides the data and the question; SAS supplies all of the compute, talent and infrastructure to answer it. SAS also has entries at less extreme points in the stack: Software-as-a-Service (SaaS), hosted analytics solutions, managed services, along with support and specific offerings based on standard public cloud standards.
Analytics in the cloud can reap the same benefits that organizations already win from other applications and infrastructure. In short, I agree with Lubbe. “I think people should stop talking about it and start doing it,” he says.
Looking for more insight? Read this whitepaper from the International Institute for Analytics on migrating analytics to the cloud.
About the Author
As the Head of Strategy and Innovation at SAS Canada Steve Holder is responsible for growing revenue and building the long-term vision for SAS in the Canada market. As a member of the SAS Canada leadership team Steve owns the SAS solution strategy, modernizing the SAS’ customer base and managing the innovation ecosystem including higher education alignment. Steve is the SAS Canada evangelist for the technology including Artificial Intelligence, Cloud and emerging technologies. Steve’s passion is making technology make sense for everyone regardless of their technical skillset. Steve can be emailed at steve.holder@sas.com. .
Recommended reading
- Fraud detection and machine learning: What you need to knowMachine learning and fraud analytics are critical components of a fraud detection toolkit. Discover what you’ll need to get started defending against fraud – from integrating supervised and unsupervised machine learning in operations to maintaining customer service.
- 6 ways big data analytics can improve insurance claims data processingWhy make analytics a part of your insurance claims data processing? Because adding analytics to the claims life cycle can deliver a measurable ROI.
- What are AI hallucinations?Separating fact from AI-generated fiction can be hard. Learn how large language models can fail and lead to AI hallucinations – and discover how to use GenAI responsibly.
Ready to subscribe to Insights now?
SAS® Viya™
Make analytics accessible to everyone and bridge the talent gap in your organization