Last week Oracle launched an upgrade to its data-warehousing technology that puts SQL, the language on which the company has more or less built its vast fortune, in the background.
After announcing the industry's "first and only self-driving cloud data warehouse," it turns out that while the user is promised a GUI drag-and-drop experience for tasks like data integration, data cleansing, and ETL, SQL is still there if developers and analysts need it.
In email correspondence with The Register, George Lumpkin, Oracle veep for product management, Autonomous Data Warehouse, said: "We have added self-service tools – and under-the-covers additional automation features – which are designed for business users. These tools do eventually execute SQL against the Autonomous Data Warehouse (ADW). For business users, we provide simple to use drag-and-drop UIs which generate SQL."
SQL now a dirty word for Oracle, at least in cloudy data warehousesREAD MORE
However, for clarity, technical users can still build queries and do their own data wrangling using SQL. "While most cloud data warehouses are built for technical users and expect all data warehouse users to understand SQL, ADW addresses both business and technical users, thereby expanding the addressable market for this offering," Lumpkin said.
Although Oracle has been one of the entrenched data-warehousing vendors for decades, Big Red's latest gambit enters a market much changed over the last five years.
As a data warehouse designed for the cloud, Snowflake has enjoyed a stratospheric rise epitomised by a $33bn IPO. Along with similar "cloud-native" offers from each of the leading hyperscalers in AWS Redshift, Azure Synapse, and GCP's BigQuery, Snowflake claims that relying on the cloud's object storage layer and separating it from compute offers speed and flexibility in previously cumbersome data warehouse deployments.
Oracle's response is to point out that its Autonomous Data Warehouse, only available on its cloud infrastructure and its cloud-at-customer on-premises offer, provides a much finer-grained specification of compute resources than Snowflake allows.
Using ADW, customers can choose the exact number of Oracle Compute Units (OCPU, equivalent to one CPU). On the other hand, Snowflake offers compute resources only in building blocks of 1, 2, 4, 8, 16, 32, 64, and 128 nodes. Similarly, AWS Redshift's latest generation of hardware offers blocks of 4, 12, and 48 virtual CPUs. Oracle claims its flexibility in pricing and provisioning allows users to pay just for what they require.
"With ADW, customers can have a 17 OCPU system and only pay for 17 OCPUs, with per-second billing," Lumpkin said.
Another difference between Oracle's approach and the cloud-native systems is the storage layer. Lumpkin said the reliance on object storage was not the same as that of the hyperscalers, but ADW was "based upon the Oracle Exadata Database Machine, which is co-engineered at the source code level with Oracle Database for peak performance, scalability and security."
The approach to concurrency is also different. Snowflake adds and drops clusters automatically to support rises and falls in degree of concurrency, which the vendor has said means customers only pay for what they use.
By contrast, Oracle allows customers to set the exact number of OCPUs which then determines the number of concurrent SQL statements the system supports. Meanwhile, workload prioritisation is built in with High, Medium, and Low settings. For example, for an ADW with 16 OCPUs, High supports up to 3 concurrent SQL statements, Medium up to 20, and Low up to 4,800, Oracle said.
If the workload outgrows the specified OCPU setting, ADW automatically provides additional OCPUs. The ADW database grows instantly, up to a capacity of three times the original OCPU setting, and the customer is charged for exactly the additional OCPUs that they use, Big Red said.
Also new to Oracle's data warehouse system is a low-code application development tool, dubbed APEX, which with its RESTful services allows other API-enabled systems to interact with the data warehouse.
Also my type...
Additionally, Oracle has extended the data types it supports. ADW supports analysis of networked relationships, though its support for graph data, at the same time, exhibits similarities to a NoSQL document store database in its support for JSON documents. All this within an on managed system.
But these features are not unique to Oracle. Teradata has been supporting graph since 2013, for example. PostgreSQL began supporting JSON files in the same year and the next open-source iteration is likely to see these features enhanced.
Henry Cook, analyst and senior director for data and analytics at Gartner, said APEX would help analytics and BI teams to redeploy their skills.
"These teams have to recruit and maintain high skilled people to make their systems run efficiently, or they spend more on the platform's to get it to perform. The idea of using AI and autonomous features is you can potentially get the best of both worlds and the feedback we get is that's pretty much true."
With SQL hidden in the background, Oracle was positioning the ADW as "kind of a platform," Cook said, offering machine learning, NoSQL data and development in one environment which would open up data warehousing to non-technical users.
Cook said Gartner had received good feedback on the ADW, and it was market perceptions that held it back against Snowflake. "It's probably more awareness than anything. Other than Oracle customers, a lot of people aren't really aware of the ADW. They do a good job for their own customers but Snowflake gets all the excitement," he said.
A far greater problem may lie in ADW only being available on Oracle's platform. With Snowflake and Teradata available across all three hyperscalers, there are options for organisations who don't want their data warehouse decision to tie them to a specific cloud infrastructure. ®