As IT industry giant HP works through a difficult multiyear rebuilding program, a rare bright spot has been its Vertica analytics platform. Like competitors IBM, Teradata and others, HP has found users ready to try column-based "data slicers" in order to speed up long-running analytical jobs that can stress conventional relational database systems (RDBMSes).
Conversations with Cerner Corp. and MZI Healthcare -- two software companies that have implemented Vertica-based analytics systems for clients in the fast-changing healthcare market -- show how column-based databases can be highly useful for those in a complex industry. Healthcare organizations turning to columnar technology and analytics say the U.S. Affordable Care Act is a major driver because it requires doctors, hospitals and health maintenance organizations to more closely measure the quality points that contribute to patient health.
For its part, McKinsey Global Institute, a well-known market research firm, found that "big-data insights" could lead to $300 billion in savings each year for the healthcare industry, if used effectively to address poorly coordinated care, fraud and abuse, and administrative and clinical inefficiencies.
Just last month, Kansas City, Mo.-based Cerner said it is in the process of moving some work from a traditional online-transaction processing (OLAP) RDBMS to the HP Vertica massively parallel processing (MPP) analytics platform. Cerner plans to use the system to load and analyze more than 10 billion performance indicators from its clients' healthcare systems. The indicators, which are used to measure how well systems are running and analyze where improvements can be made, include data about CPU and memory usage, system speeds, application usage statistics, and much more. The goal is to ensure stable services delivery and streamline operations, according to Cerner.
The performance indicators are part of a Real Time Measurement System embedded in Cerner's Millennium software. It helps the company tune operations performance while letting customers "look into" the system to see statistics on job execution, said Dan Woicke, director for enterprise system management at Cerner.
Before going live with Vertica, Cerner's RDBMS-based performance monitors would send alarms when, for example, a client's memory use reached a key threshold. But the earlier system didn't retain a record of the data points that led up to that warning, Woicke said. That has all changed now. When warnings are sent, Cerner can see precisely how the problem escalated.
Cerner's Real Time Measurement System can be used to provide feedback on how physicians and others are using their system, which can help improve healthcare efficiency.
"I can tell what a physician does [with the system]. That is very valuable to me. For example, if they are not using the built-in time-saving features, we can work with them to use the system correctly," Woike said, adding that physicians "don't want to be [buried] in the application -- they want to be working with patients."
Cerner is also using the Vertica platform as part of a new Health Facts Reporting service that will analyze data generated by health systems to create information intended to improve the quality of patient care.
Healthcare organizations considering the purchase of an analytics platform should keep three key criteria in mind, according to Woicke. They include "one, speed and performance of queries on common data sets; two, concurrent support for up to 400 clients issuing SQL commands at the same time; and, three, price."
Affordable care slicing and dicing
Healthcare software and services provider MZI Healthcare is using the Vertica platform as a repository of patient information within MZI's EZ Analytics on the Cloud product, according to Greg Gootee, a product manager with the Valencia, Calif.-based company. Column-based Vertica "enables more flexible and faster data loading," he said.
The needs for faster and more complex processing have grown as new regulations seek to improve the overall quality and efficiency of healthcare systems, Gootee said, adding that the Affordable Care Act causes healthcare providers to more closely measure the effectiveness and cost of patient care. "Today, the structure of the data is changing rapidly," he said. "Vertica, with its column structure, has made a big difference."
Gootee explained that prior to Vertica, MZI's IT team would regularly have to create OLAP cubes to manage, process and load important data. "Today we pretty much drop the data in, with some cleaning, and work on it quite rapidly," he said. "It used to take a day or longer to load [some jobs] that we can now do in minutes."
Gootee also credited the software for its support of standard SQL interfaces. MZI developers can work with the system as long as they know SQL, he said, adding, "We needed to be able to slice and dice data in a million different ways."
Analytical engines 'only growing'
Lately, good news has been sparse for HP, but Vertica appears to be an exception. Beleaguered HP CEO Meg Whitman last month disclosed that Vertica saw "triple-digit" growth in the fiscal quarter just ended. HP has been promoting Vertica, purchased with Vertica Systems Inc. in 2011, along with other software as part of other data center and services programs.
Still, the column-based approach is not a "silver bullet" in all cases. And established relational databases are beginning to take on column-based attributes that may suffice for many shops' needs. "Vertica has been a very good acquisition for HP. It is a rare gem of growth across the company," said Dana Gardner, president and principal analyst at Interarbor Solutions. "With HP, the original Vertica people are finding a ready sales channel, and a global one at that."
Gardner said the industry is seeing demand for high-performance analytical engines "only growing."
HP's Vertica's MPP engine is often inserted in place of established RDBMSes that may falter in the face of larger jobs and more ad hoc queries, he said.
"For customers that I have spoken with, it's about replacing a relational database system. They need very fast queries, and cost is an issue. Where data is large and velocity is fast, they find the RDBMSes are slow and they can be expensive," Gardner said.