- Oracle includes support for anomaly detection, recommendation, and multivariate time series forecasting in the engine along with some well-known machine learning apps.
- For efficient monitoring and workload planning, ML can be integrated with MySQL Autopilot.
Oracle Corp. enhanced the MySQL HeatWave product with latest automation features, ML capabilities, and enhanced performance on the cloud of Amazon Web Services Inc., capitalizing on its growing market share for hybrid transactional/analytical database management systems.
HeatWave, an in-memory analytical accelerator that was unveiled in late 2020, expands to thousands of cores, supports both real-time analytics and online transaction processing, and has made waves with its outstanding price-performance. Developers don’t need to extract data from another database to fill training models thanks to the embedded machine learning features.
As compared to AWS’ SageMaker managed machine learning service, which is priced independently from AWS’ Redshift cloud data warehouse, training is offered free for customers. According to Oracle, MySQL HeatWave is up to 99% less expensive and 25 times quicker than Redshift.
Oracle includes support for anomaly detection, recommendation, and multivariate time series forecasting in the engine along with some well-known machine learning apps. Anomaly detection is widely used in financial services to watch machinery and manufacturing environments as well as to find out why a credit card is being declined. Customers or viewers of videos can receive personalized recommendations from recommendation algorithms. Using numerous variables, multivariate time-series forecasting can resolve complex issues.
HeatWave AutoML can integrate models based on the Open Neural Network Exchange standard without the need for a machine learning framework. Nipun Agarwal, Sr. Vice President for MySQL database and HeatWave at Oracle, added that a large portion of the training procedure was also automated by the company.
He mentioned, “Not only do we now offer machine learning processing inside the database, but the training process, which is the more complicated part of machine learning, is fully automated as well.” This involves a voluntary choice of features, algorithms, and parameters. Oracle stated that other cloud services suggest only algorithms that require users to choose the most suitable one and tune it manually.
Agarwal reported, “All models can be explained because we use model diagnostic techniques. Users can run explanations directly from the console, which means no ML expertise is required. We have also introduced this notion of what-if scenarios where the user can toggle various attributes to see if it changes the outcome of the machine learning model.”
For efficient monitoring and workload planning, ML can be integrated with MySQL Autopilot. The feasibility of OLTP workloads provided by the Auto Shape Prediction can “make it very easy to see why autopilot is making the decisions does,” said Agarwal. Moreover, it recommends tables that can be purged or unloaded from the memory to lower the pricing. The suggestions are backed by visual analytics of archived performance trends that include buffer pool hit rate and throughput.
Oracle is also enhancing the integration of S3 object storage, CloudWatch, and PrivateLink into HeatWave to optimize it for use on the Amazon cloud. For mixed columnar representation, HeatWave now offers an optimized storage layer based on S3 object storage.
HeatWave copies the data it receives from MySQL to its scale-out data management component, which is based on S3. If reloading is necessary, Oracle said, the data can be loaded without the requirement for transformation, leading to noticeably quicker availability and recovery periods. Data never departs the AWS cloud, so egress fees are not applicable.
Agarwal commented that HeatWave on AWS works 10X faster than Snowflake Inc.’s namesake data warehouse and 20X as compared to Redshift. “HeatWave is designed from the ground up for distributed scale-out query processing. It has been designed for commodity hardware and since we use machine learning-based automation the system learns and improves on the fly,” he added.
Finally, Oracle is extending HeatWave nodes and altering their design and cost. For an additional USD 16 per month, a 32-gigabyte shape will be added to the current 512-gigabyte option. A standard 512-gigabyte node can now handle up to 1 terabyte of data, up from its previous limit of 800 gigabytes. According to Oracle, this results in a total price performance improvement of 15%.