Asics Digital, the data-centric subsidiary of Japanese sportswear giant Asics, recently shifted its data warehouse from a creaking Amazon Redshift instance to Snowflake's cloud-native solution in a bid to better support its growing portfolio of digital services, such as the Runkeeper and Asics Studio apps.
The 70-year-old Japanese brand acquired the GPS fitness-tracking company Runkeeper in 2016, forming part of a digital strategy that mirrors that of its competitors like Adidas, UnderArmour and Nike, all of which offer a run tracking application of their own.
Combine this with the homegrown Asics Studio app – a subscription service which provides users with a range of curated workouts – and the company clearly sees value added digital products as part of its future direction in the lucrative health and fitness space, which is estimated to be worth £5 billion in the UK alone.
How it got there
Back in 2015 Asics was running much of its business on Amazon's Redshift data warehouse, but as its data volumes continued to rise the company was quickly running out of storage.
"We had to keep taking clusters down for maintenance to increase the size or clean out the data we thought we weren't going to use," Chris Druin, manager of advanced analytics at Asics Digital explained to Computerworld.
Snowflake was built to offer customers almost limitless scale and concurrency by effectively spinning up new cloud instances (S3 on AWS, for example) for each workload to effectively run as a standalone data warehouse but all under the same roof, so data-science queries never tread on the toes of BI, for example.
Druin admits that the "migration wasn't perfectly simple, but as Redshift and Snowflake primarily source data from Amazon S3, for some parts of our infrastructure it was as simple as getting Snowflake to start picking that data up.
"In other cases we changed some load processes, because Snowflake gave simple and flexible options there, such as handling semi-structured data. We have a fair bit of information that comes in JSON blobs and Snowflake can read those natively as a data type and access them with near-native performance levels."
Asics moved its first set of production data to Snowflake in 2016 and ran both systems in parallel until the organisation was confident enough to commit to a full migration.
What are the benefits?
Before Snowflake, Druin said employees could be waiting in a virtual queue for queries to end due to the lack of ability to run concurrently.
"Snowflake solved a lot of those problems for us by decoupling storage and compute," he said.
The company started by spinning up a Snowflake instance for its extract, transform, load (ETL) warehouse, which is relatively small but always on. Next was the business intelligence (BI) warehouse for reporting, which powers a set of internal Tableau dashboards.
The biggest benefit so far from a pure operations point of view is the reduced maintenance obligations on the IT team at Asics.
"We spent a lot of time on maintenance with our previous solution," Druin said, "so not having to schedule periods of downtime and subsequently backfilling, or making good on missing data that would have come in, has been a huge benefit."
Naturally this has cost and productivity implications, such as saving a member of the team from overseeing a maintenance period scheduled for a Saturday morning.
By running workloads in their own 'standalone' data warehouses, data scientists at Asics are now able to run ad-hoc queries without stepping on those operational workloads.
The Asics Digital business unit now stores 17 terabytes of data with Snowflake, including data points from 50 million users across 180 countries, from GPS run information to buying history.
"We have been trying to build what we call a single view of the customer and that was more challenging in the past when we didn't have all of our data in one place," Druin said. "Now that we do, that enables us to quickly spin up an instance and see the history of our relationship with you as a customer."
It also allows Asics to start delivering the accumulated knowledge it has on runners back to them in the form of useful insights. For example, incorporating third-party weather data into the Runkeeper app to nudge users to run at a different time, or consider an alternative workout if a storm is on the horizon.
Asics isn't the first fitness company to turn to Snowflake for this sort of thing either. Computerworld spoke to Strava earlier this year about how it was running queries overnight while its Redshift instance creaked under the weight of its growing data volumes.The San Francsico-based company stores more than 120TB in Snowflake today.
Strava is also using Snowflake to allow its data scientists to produce things like its global Heatmap or to quantify effort through heart rate, or optimise its Grade Adjusted Pace metric.
This is the sort of thing that makes Druin most excited. Now Snowflake can do much of the heavy lifting on the backend, the advanced analytics team at Asics can start to "find new sets of data to bring to customers and provide insights into their workout and gear performance," he said.
"Those are the things on our radar. There will be new things in the fitness apps like Runkeeper and Asics Studio to power new features and functionality in those apps," he teased.