For one of the world’s largest medical device companies, innovation isn’t just about competitive differentiation, or even survival. It’s who they are. Their story began with the invention of a device that revolutionized cardiac care. Now their therapies help alleviate pain, restore health, and extend life for many millions all around the globe, with products spanning cardiac devices, surgical tools, cranial and spine robotics, insulin pumps, and more.
Today, innovation remains at the heart of the company. But innovation depends on data — which is why, once they understood that their data platform was holding them back, they knew they had to act.
They had long been collecting Internet of Things (IoT) data from their various devices but struggled to properly equip the right teams with that data. In recent years, the company sought to experiment with more sophisticated predictive analytics, such as identifying when a device might become likely to fail. But enabling these experiments meant allowing a plethora of siloed teams to all begin spinning up their own ad hoc data projects. And on their single-tenant, on-premises data warehouse, this quickly exceeded the limits the existing infrastructure could handle.
As the requests poured in, dozens, if not hundreds, of analytics projects were spun up. But because individual teams had little awareness of how the performance demands of their own workloads were adversely impacting another’s, “noisy neighbor” issues were a problem almost immediately. And before long, the available storage and compute was gobbled up, which led to a crisis because their shared environments required an ever-increasing amount of licensing and infrastructure to meet demand.
To complicate matters, a chargeback model had not been implemented, meaning infrastructure expenses were all being charged back to central IT with no way of tracking how much each team was responsible for. They would never be able to support the resulting cost structure and continual demand for resources without hiring more administrators, requiring long nights of troubleshooting increasingly complex problems.
Clearly, the current model was unsustainable. To continue their tradition of innovation, the company needed to re-evaluate their current model and find a different approach to meet the growing needs of their internal customers. They needed a new, cloud-based central data platform that could integrate with existing Business Intelligence tooling.
After some initial exploration, the company settled on the Snowflake Data Platform because it integrated easily with their existing tools, was easy to define a chargeback model, and allowed them to scale their business as demand grew. However, implementing an enterprise-quality ecosystem for data pipelines — especially for cost- and performance-optimized data utilization — presented its own challenges:
To overcome these challenges, the company knew they needed access to expertise, processes, and standards, both around data analytics in general, and the Snowflake platform in particular.
Ultimately, they chose to partner with phData. The device company was confident, following a successful proof-of-concept project, that they could rely on phData’s track record of successfully delivering the automation and expert services required to build enterprise data analytics solutions on Snowflake and AWS.
The phData team provided a range of services in order to prepare the medical device company to fully migrate from their Oracle data warehouse to Snowflake and AWS. This included:
The initial priority was to translate the company’s existing business processes to function with Snowflake, and take full advantage of the platform’s cloud-native architecture. This included setting up a framework for future migration efforts, defining workspace parameters within Snowflake, and creating rules for handling chargebacks and maintaining budget transparency. As a result, individual business units would now be responsible for the costs of their own projects, rather than dumping everything on central IT.
Because the company needed to support such a massive volume of potentially ad hoc projects coming in from so many different teams, streamlining the process of creating Snowflake workspaces and onboarding new uses was absolutely critical.
To achieve this, phData provided them with their proprietary automation tool, Tram, and helped them configure it for future use. Onboarding processes that would have involved 1,000s of SQL commands could now be done automatically, reliably, and repeatably. As a result, end-users can now consistently get projects and experiments up and running on Snowflake in a matter of hours, rather than days or weeks.
By relying on the monitoring tools and services from phData Cloud DataOps, the company is able to identify poorly optimized queries, clustering, and credit usage issues. This ensures the right warehouse sizes for the right workloads, which is critical to minimizing costs.
For example, queries that are piling up or taking too long may require a larger warehouse size to support more concurrent users, whereas larger warehouses that aren’t being fully utilized may need to be downsized.
phData was initially brought in to help the device company configure Snowflake according to security best practices by providing guidance around creating Secure Views and configuring row-by-row security access. But, phData went above and beyond that scope because of the new and unexpected challenges brought on by the COVID-19 pandemic.
With nearly all the device company’s employees working remotely through a VPN, minor inconsistencies with their VPN providers’ assignment of IP addresses wound up creating major headaches. phData took on the problem, working with the third-party provider to resolve the issue.
With the framework of business processes, solution architectures, and automation tools provided by phData, the medical device company has everything they need to support innovation with Snowflake and AWS.
Learn how phData can help solve your most challenging data analytics and machine learning problems.