Open MySQL databases are the workhorse of data processing and analytics, and currently hold a 43% share of the relational database market.
But while traditional MySQL databases offer reliability and cost-savings, they suffer from limited reporting and analytics capabilities, low automation potential and zero machine learning ability. Database administrators have overcome this by adding on separate services in a “right tool for the right job” approach. It’s a method certainly employed by other cloud giants, where a smorgasbord of database options is fast becoming the norm.
Oracle Corp. chose a different path. With the popularity of its market-leading Oracle Database showing a very slow, but steady, decline as cloud databases started to pick up share, the company challenged convention by converging all database functionality in its MySQL HeatWave fully managed database service.
“We see Oracle, which is taking the Swiss Army Knife approach, converging database functionality, enabling analytic and transactional workloads to run in the same data store, eliminating the need to ETL, at the same time adding capabilities into its platform like automation and machine learning,” stated theCUBE industry analyst Dave Vellante in a CUBE Power Panel on the database market.
Vellante dug into MySQL Heatwave’s rapid evolution during a CUBE Conversation this week with Nipun Agarwal (pictured, left), senior vice president of MySQL Database & HeatWave at Oracle, and Kumaran Siva (pictured, right), corporate vice president of strategic business development at Advanced Micro Devices Inc. They discussed how in-database machine learning is becoming essential to provide the performance levels that modern analytics demand.
Automated training removes need for ML specialists
There are multiple reasons that Oracle chose to integrate ML into MySQL HeatWave rather than take the specialized database approach chosen by AWS and other competitors, according to Agarwal. First is security.
“Customers don’t need to move the data. And if they don’t need to move the data, it is more secure because it’s protected by the same access-controlled mechanisms as the rest of the data,” he explained.
In addition, a single database makes management easier, as customers don’t have to deal with multiple services. It’s cheaper, as customers don’t have to pay for external ML services, and it’s faster. Another differentiator that saves money and makes customers’ lives easier is that Oracle has automated the ML training process within MySQL Heatwave. This removes the need to hire ML experts, as no specific parameters need to be provided, just the source data and the task on which the ML is to train.
“This is something which is very important to database users, very important to MySQL users,” Agarwal said. “They don’t really want to hire data scientists or specialists for doing training.”
Automating ML training speeds up the process, which makes more frequent retraining possible and enables models to be kept up-to-date.
“And, as a result of the models being up to date, the accuracy of the prediction is high,” Agarwal added.
Customers can then run inference to obtain actionable insights from the models. And, in what Agarwal said is “the most sought-after request,” they can obtain explanations for any model generated or trained by MySQL HeatWave ML.
How does MySQL HeatWave ML beat the competition?
MySQL HeatWave’s impressive performance statistics were discussed by Agarwal and Siva in a previous CUBE Conversation. And when it comes to MySQL HeatWave’s ML performance, Oracle provides numbers that come with bragging rights. Citing a comparison between MySQL HeatWave and AWS Redshift ML, the aggregate performance of HeatWave ML on 12 data sets for classification and six data sets on regression was 25 times faster than Redshift ML at 1% of the cost, according to Agarwal.
Research conducted for Wikibon, SiliconANGLE Media’s sister market research firm, by analyst Marc Staimer put MySQL HeatWave ML up against the competition. Results showed that in the total cost of ownership/performance, MySQL HeatWave beat Amazon Redshift ML by 82x, Azure SynapseML by 85x and GCP BigQuery ML by 88x.
“There are many techniques that we have developed specifically for machine learning,” Agarwal stated. “[These] give us the better performance, better price performance, and also better scalability.”
This performance comes not from the more powerful Graphics Processing Units, as most would assume, by the less-energy intensive Central Processing Units. This is enabled by AMD’s EPYC processors that power the Oracle Cloud Infrastructure.
“What you see in the AMD architecture for EPYC for this use case is the balance and the fact that you are able to do the pre-processing, the AI, and then the post-processing all seamlessly together — that has a huge value,” Siva said.
With MySQL HeatWave ML, inference, explanation and training all use the CPU in the same OCI infrastructure. Using the same infrastructure results in better TCO, as well as better performance because “you’re bringing the data to the computation,” according to Siva.
“The in-database HeatWave ML puts Redshift ML and Snowflake on notice,” Ron Westfall, senior analyst and research director at Futurum, said in the CUBE Power Panel. “Are these solutions more like yesterday’s tech in terms of engineering, performance and cost? Because they’re slower and more expensive, chances are the answer is yes.”
Here’s the complete video interview, one of many CUBE Conversations from SiliconANGLE and theCUBE, and be sure to check out more of SiliconANGLE’s and theCUBE’s coverage of enterprise technology, digital transformation, and cultures of innovation: