Database News from Oracle, Pinecone, SingleStore, Promethium
Something for everyone: in-database ML; startup funding; new hires; & a partnership with Snowflake
It’s a jam-packed day in the cloud database market led by Oracle’s announcement of in-database machine learning for MySQL HeatWave, with other news from Pinecone Systems, SingleStore, and Promethium.
Oracle Adds In-Database ML to MySQL HeatWave
Let’s start with the big dog. Oracle introduced in-database ML for MySQL HeatWave, the in-memory query accelerator for its MySQL cloud service.
As the name implies, the new in-database capability makes it possible to run ML models within the database rather than externally. The advantages include efficiency, performance, cost, security, and ease-of-use, according to Oracle.
Edward Screven, Oracle’s Chief Corporate Architect, said that the challenges of adopting ML — training algorithms, generating models, running inference on production data — have prevented many organizations from using it.
“Why is it that more people don’t use machine learning? It’s very hard to actually use,” he said. MySQL HeatWave ML automates and simplifies otherwise manual ML processes. HeatWave ML is available now at no extra cost to HeatWave customers.
In related news, Oracle rolled out other HeatWave updates, including the ability to automatically scale up & down clusters to any number of processor nodes, 2X data compression per node, and a “pause and resume” feature for occasional usage and cost savings.
I was briefed on HeatWave’s new capabilities by Nipun Agarwal, Oracle SVP of MySQL and HeatWave. One of the most interesting new capabilities in the ability to explain ML models, which adds an important governance component for things such as regulatory compliance, repeatability, and causality.
“Explanations are a very important part for enterprises,” says Agarwal. “Why was a credit card denied? Why was a loan denied? Enterprises have an obligation to ensure there is no bias. There’s an implicit right for explanation.”
Oracle also touted new TPC-DS benchmark results that demonstrate price/performance advantages of HeatWave vis-a-vis Snowflake, Amazon Redshift, Microsoft Azure Synapse, and Google BigQuery.
These are the latest round of tweaks to Oracle MySQL, which can be used for OLTP, OLAP, and now in-database ML workloads. Last August, Oracle announced an “autopilot” capability for HeatWave that uses ML to automate some aspects of DB administration, including provisioning, parallel loading, data placement, query scheduling, and more.
MySQL HeatWave is available now with Oracle Cloud Infrastructure. On Oracle’s Q3 FY22 call a few weeks ago, Chairman & CTO Larry Ellison said HeatWave is, or will become, available on AWS and Azure, as well. “So it is a multi-cloud product,” Ellison said.
Pinecone Secures $28M Series A Funding
In other news, Pinecone Systems announced that it has secured $28 million in Series A funding to continue developing its vector database.
As I have previously reported, Pinecone’s vector database provides similarity search as a cloud service. Use cases include recommendations, personalization, image search, and deduplication of records.
A vector, or vector embedding, is a string of numbers that represents documents, images, or other data. Vectors are used in the development of machine learning apps. A vector database stores, searches, and retrieves the representations by similarity or by relevance.
Pinecone exited stealth mode in January 2021 with $10 million in seed funding. With this latest round of investment, the company is doubling down on AI-powered search capabilities that go beyond keyword search. “Search is in desperate need of modernization,” said Pinecone founder and CEO Edo Liberty.
Pinecone describes itself as a search infrastructure company. Its technology enables storing and searching through AI-generated representations of content in machine-readable format. Developers can use it to build search apps “that anticipate and understand” user needs, and not simply match keywords.
Below is my podcast interview with Pinecone CEO Edo Liberty shortly after the company’s launch last year.
SingleStore Hires Two Microsoft Veterans
SingleStore has hired two Microsoft veterans to lead database development and product management.
Shireesh Thota joins SingleStore as senior VP of engineering to oversee development of the company’s multi-model SQL database. Thota worked in various roles at Microsoft for 15+ years, including running engineering for Cosmos DB and Postgres Hyperscale (Citus) services.
In addition, Yatharth Gupta will head up product management and design as VP of product management. Before joining SingleStore, Gupta worked for 14+ years at Microsoft, where he most recently led the product team for Azure Databricks products.
Both Thota and Gupta will report to CEO Raj Verma. Watch for my upcoming podcast interview with Raj Verma in the days ahead.
Promethium Acts Flaky
And finally, Promethium says it will deliver no-code “federated query pipelines” for Snowflake’s Data Cloud.
Promethium describes its technology as a data fabric solution, incorporating a data catalog and more than 200 data connectors to sources including AWS S3, Salesforce, SAP, and other database environments.
The idea is to use Promethium to move data to Snowflake’s cloud faster and without the complexity sometimes associated with ETL and data migration. A Promethium Free edition for Snowflake is available.
I wrote about Promethium a few weeks ago on the Acceleration Economy website. You can see the article here, “Three New Analytics Products Bring Fast and Flexible Access to Far-Flung Data.”