Machine learning is delivering real profits to more companies by automating decision-making and enhancing digital products. Yet the flood of data pouring into these statistical prediction systems has created another problem: How to get reams of data ready for powerful computers to analyze amid of a dearth of machine learning expertise.
Twenty-seven percent of companies surveyed by McKinsey & Company last year reported that at least 5% of their operating profit was attributable to artificial intelligence (AI) usage, up from 22% of companies in 2020. In the survey of 1,843 global companies, 56% said they’re using AI in at least one area, chiefly to automate customer service, improve products, segment customers, and model risk.
Yet organizations have their data hived away in myriad databases with different departments using different approaches. Ferrying information to machine learning systems and putting it in a form that’s ready to analyze adds cost and time drag. “AI needs data, and for enterprises, moving data around is cumbersome, slow, and expensive,” says Holger Mueller, vice president and principal analyst at tech advisory firm Constellation Research.
The latest version of Oracle MySQL offers a better approach: Build machine learning models into the database, so you can leave the data where it is. “Bringing the machine learning to the data is a smart architecture and design choice,” Mueller says.
First released two years ago, MySQL HeatWave runs as a service on Oracle Cloud Infrastructure and lets users of the widely adopted MySQL Database combine transaction processing and data analysis in a single database. It runs database queries in computer memory to increase their speed.
Machine learning built in
The latest iteration of the software puts machine learning capabilities inside the MySQL HeatWave Database and can automate the process of creating and training ML models, then unleashing them on live data. MySQL HeatWave can train models on available data and tune them to improve accuracy based on how well they predict outcomes. That means when organizations’ data is inside MySQL Database, as much of it already is, they can skip the work of extracting it, changing its format, and loading it into another ML system—a process that can take hours. Developers and data analysts can also build ML models using the familiar database SQL language without having to learn new tools or languages.
“HeatWave is easy to use, period,” says Matt Kimball, an analyst at Moor Insights & Strategy. “The biggest barrier to adoption in the world of ML is complexity—organizations can’t find the right talent and the best hardware. With HeatWave, they can realize the promise of machine learning more quickly.”
Sales of cloud data warehouses are expected to grow sharply this year as businesses use them for tasks, such as managing supply chains, hunting for sales opportunities, or analyzing medical data. The systems tap databases, the web, business applications, and sensors to gather data in one place, analyze it, and give companies a clearer view of operations, sales, and customer behavior.
Oracle published machine learning benchmarks performed across a large number of publicly available machine learning datasets using different cloud providers—research noted by several analysts as an industry first. On average, on the smallest cluster, MySQL HeatWave ML-trained machine learning models were 25 times faster at 1% of the cost of Amazon Web Services’ (AWS) Redshift ML. The benchmarks based on public ML datasets show MySQL HeatWave delivering better price performance than Google’s BigQuery, Microsoft Azure Synapse, and Snowflake, as well as AWS’s Redshift with its query accelerator.
Less time spent using a cloud service can mean lower computing bills, and the new MySQL HeatWave version also lets customers scale up without downtime.
Last year, Oracle introduced software called My SQL Autopilot for using AI to optimize and automate the MySQL HeatWave service. “Autopilot used machine learning to operate the database. Now it’s about using ML on the data,” says Mueller.
MySQL HeatWave can also take models tuned in MySQL and use them to answer queries from popular machine learning notebook services for code and data, such as Jupyter and Apache Zeppelin.
Oracle plans to release MySQL HeatWave to run on other cloud providers’ infrastructure, including AWS, Oracle Executive Chairman and Chief Technology Officer Larry Ellison said during a conference call with financial analysts last month.
Edward Screven, Oracle’s chief corporate architect, said during an online presentation that companies are finding it hard to achieve their machine learning ambitions. “If you use conventional tools, if you use what comes out of the box from most vendors, from most clouds, machine learning is hard, and it's expensive,” he said. In contrast, “you don’t need to be an expert to use machine learning with your data stored in MySQL HeatWave.”
The new capabilities are available at no extra charge for MySQL HeatWave customers in all of Oracle’s global cloud regions. Oracle is also making both the ML and TPC-DS benchmark tests public for customers to perform themselves. “That really lends credibility to the price and price-performance claims,” says Moore Insights’ Kimball.
Source: oracle.com
Posts
0 comments:
Post a Comment