Medaxion has found a solution to the analytics problems that plague companies across the health care sector and beyond. John Toups, CTO of Medaxion, puts it plainly. “By combining the ease of presentation and abstraction of Looker for analytics, with the technical prowess of SingleStore as the database behind Looker, we now have simply the best analytics in healthcare.”
Medaxion has changed the working lives of anesthesiologists, its core customers. These highly skilled professionals must deliver absolutely critical medical care, day after day, while also meeting demanding business requirements. These men and women, who once generated a blizzard of paper at every turn, now “record their anesthetics in Medaxion Pulse,” according to Toups, “and through Looker and SingleStore, that data is now actionable.”
Today, surgical teams and patients who sometimes waited hours for an anesthesiologist to arrive are now helped much faster, with the aid of predictive analytics. “If it’s cold and snowy, some elective surgeries might get cancelled because it’s hard to get around,” Toups shared. “At the same time, ER admissions might spike because of hazards on the road and the outdoors. We can monitor the data, minute by minute, and help put anesthesiologists where they’re needed most” – getting care to more people, faster.
Looker Exposes a Slow Database
Looker is an analytics tool that’s designed to make analytics fast and easy. Looker integrates deeply with multiple data sources, rapidly converting user requests to live results. Looker also provides an abstraction layer, LookML, a modeling language that keeps users from having to learn SQL or other code.
With Looker’s speed, and the ease of use provided by LookML, demand for analytics at Medaxion rose sharply. But increased demand meant more queries coming into Looker. And, while Looker itself could handle the volume, MySQL – Medaxion’s previous underlying database – couldn’t keep up.
SingleStore Solves the Problem
Toups and Medaxion had a problem. How did they solve it? Simple: by leaving MySQL and moving to SingleStore.
“Much of what my customers want from analytics is real-time operational information, and there is enormous interest in monitoring and improving quality of care,” said Toups. For instance, anesthesiologists need to be near patients who will be needing surgery. It’s a dispatching problem that is, in a way, similar to the issue faced by Lyft and Uber: have the right person, providing the right service, in the right place, at the right time. But for Medaxion’s anesthesiologist clients, and the patients who need them, the stakes are higher.
With Looker running on MySQL, Medaxion experienced significant problems. “At first, our analytical reporting was incredibly slow,” said Toups.
“When we first started implementing Looker, a couple of years ago, we did a traditional ETL ‘lift and load’ into a MySQL reporting warehouse,” reported Toups. “This resulted in about 600GB of data. Now, on SingleStore, I use columnstore with compression and SingleStore’s native sharding.”
“The underlying data includes measurements such as systolic and diastolic blood pressure, along with heartbeat and respiration and a variety of medical history information, all of which are strongly correlated to one another,” Toups continued. “I use associative techniques to compress that 600GB down to a 20GB dataset. Simply unprecedented compression. We always had similarity on our data, but we couldn’t take advantage of it in the MySQL environment.”
“It was taking 30 or 40 minutes to generate retrospective quality data on the MySQL platform,” Toups said. “Because we couldn’t easily cache the entire data set in memory on MySQL, it was constant disk thrashing. But on SingleStore, the same analysis runs in less than a minute, and most queries return in under a second.”
Toups summed up Medaxion’s progress: “We replaced the middle of the analytics chain with SingleStore. Looker is the body of the rocket ship that carries the information my customers – and their patients – need. SingleStore is the engine of that rocket ship.”
Medaxion saw dramatic performance improvements with the move to SingleStore.
Re-Plumbing the SingleStore+Looker Solution
Solving complicated data flow problems is not always easy. It took a certain amount of expertise and effort to find the right solution to specific problems.
Medaxion wanted a solution that would be easy to implement. It wanted as little change as possible in its pre-existing processing architecture. It didn’t want to have architects working for months, nor did it want to end up with a big team of operations people to monitor and manage a complicated solution.
SingleStore frees Medaxion from a lot of operational overhead, while empowering the anesthesiologist users. The technical people at Medaxion also appreciate the architectural purity of SingleStore. “Scale can be difficult,” said Toups. “But scale done well is a joy.”
New Architecture Delivers Results
With SingleStore and Looker, each day sees new operational efficiencies. “Installing and maintaining SingleStore has been a no-effort proposition,” commented Toups. “And, because the previous solution was so slow, we used to do a lot of efficiency programming, indexing, and query planning. With SingleStore, I’ll pre-aggregate when there’s an obvious need. But I can also just let our customers do as they wish with the data.”
The results are impressive. “We’ve worked to reduce the elapsed time to less than 30 seconds between a data event and a reportable fact,” said Toups. “This is an enormous improvement, and we achieved it without much change to the underlying architecture.”
Medaxion Deploys Singlestore Helios
Medaxion has continued to innovate in its use of SingleStore, serving as an early adopter of Singlestore Helios, SingleStore’s elastic cloud service. Medaxion served as a design partner for the new offering, contributing feedback and ideas as they stood their instance up on Singlestore Helios and moved to production.
“We get all of the advantages of SingleStore, with no increase in cost relative to hosting it ourselves in the public cloud, but tremendous advantages in operational savings and simplicity for the team. We save time and effort on managing our deployment, and SingleStore’s people can often fix issues before we’re even aware there might be a problem. We’ve helped make their product better, and they’ve helped make us even savvier users of their product.”
Looking Forward
It’s only been six months since Medaxion became a SingleStore customer. Medaxion’s rapid progress has fundamentally altered Medaxion’s future and the future of its anesthesiologist customers.
SIS
And Medaxion is just getting started. By deploying SingleStore and Looker together, Medaxion has enabled a data culture amongst its customers. Now, the anesthesiologists can do their own data mining, with results that are already amazing.
Medaxion is building up a huge base of data to be mined going forward. Predictive analytics, machine learning, and AI can help Medaxion, and its empowered customers, to improve both business practice and medical practice around the crucial discipline of anesthesiology.
Third, there may still be room for improvement in Medaxion’s data architecture. Toups and others at Medaxion are actively investigating the potential for far-reaching change in how Medaxion operates.
SingleStore has helped Medaxion to save time, save money, and dramatically improve outcomes in the short term. At the same time, Medaxion is opening the door to even bigger, and better, changes in the future.
