Oracle Cloud’s 2020 strategy will heavily focus on addressing customers’ data sovereignty high-availability requirements by nearly doubling cloud center regions and redundancy; while its autonomous database grows into more of a multi-functional umbrella offering.
An example is how Oracle has extended elasticity from compute to storage. It is virtualizing its storage into block volumes that can be configured by performance or cost. While in other clouds, customers typically select specific tiers of storage, such as disk, SSD, or memory, in Oracle Cloud you can buy performance units on the fly that can increase or decrease IOPS. While currently this is a manual process, Oracle plans to add an automated option that would, in essence, extend autoscaling to storage.
It's all part of the narrative that Oracle is spelling out, that the other folks did cloud first and they had a chance to learn the lessons of the first generation. With the Azure partnership, there is another part of that narrative: Oracle is not trying to be all things to all people. The tie-in with Azure reflects the fact that most Oracle customers already have Microsoft in the front office.
Growing the footprint
A year ago, we wrote that Oracle Cloud was preparing for global build-out. With the Gen2 infrastructure introduced at OpenWorld 2018, the template was established for expanding the footprint. A year ago, Oracle counted less than a half dozen global regions, today it is up to 20, with plans to almost double that by the end of the year.
For some perspective, Oracle uses the term "regions" differently from AWS or Azure – its regions are the equivalent of data centers or availability zones with the other clouds. And so, a region with AWS or Azure means that there are at least two separate data centers in it (and increasingly, three). So, most of Oracle's early regions have only one data center, but this year, that's going to change. About half the growth that Oracle is planning in 2020 will be adding those additional data centers in the geographies where it already sits. Today, Oracle Cloud has in-country disaster recovery in the US and Japan; by year's end, it plans to have nine more countries with duplicate data centers that are based in different cities. A key driver in all this is responding to data sovereignty regulations requiring companies to keep their data within the country of origin.
As Oracle raises the profile of its cloud positioning, it is being realistic in not trying to be all things to all people. While it contends that its Infrastructure-as-a-Service offering is based on more current technology than rival clouds, the primary role of IaaS at Oracle Cloud is to serve as building block to the company's enterprise application SaaS and autonomous database PaaS services, where it better differentiates with the usual suspects. The partnership with Microsoft, for building high-speed links and unified identity and access management between Oracle and the Azure clouds, reflects the fact that Microsoft, not Oracle, is the default front office.
The autonomous database to become more than database
The autonomous database is a key differentiator for the Oracle Cloud. As we noted last fall, with the autonomous database, Oracle now has at least a year or two track record with some of the earliest clients, with the common themes being superior performance at lower cost, and changing of the DBA role.
Going forward, Oracle's Autonomous Data Warehouse will spread its functional footprint to encompass data transformation/ETL with a no-code drag and drop experience that will be useful for relatively simple data transformations (it won't replace Oracle Data Integrator for the more complex work performed by data engineers). Additionally, it will have an "Auto Insights" capability for data discovery on incoming data, for detecting outliers and relationships with existing data. And there will be a capability for performing machine learning inside the database geared to "citizen data scientists." It will automate the selection of algorithms, feature selection, and parameter tuning. Rounding it off is Oracle visual database application development language (APEX) and federated query capability via Cloud SQL (not to be confused with the Google Cloud service of the same name).
Oracle is hardly the only analytics provider to expand the role and definition of the cloud data warehouse to go beyond just being a database. Last fall, Microsoft unveiled Azure Synapse Analytics, which embeds the data pipelining capabilities of Azure Data Factory into the data warehouse. At the other end of the spectrum, SAP extended its HANA Data Warehouse Cloud, but with self-service analytics capabilities adapted from its Analytics Cloud. All of these signify an emerging trend among cloud SaaS and PaaS providers to extend the database into a one-stop shop for either the data engineer or business user.
Following the 2018 acquisition of DataScience.com, Oracle has been refashioning the platform to become a native service in the Oracle cloud. The service is focused on the lifecycle management of data science and machine learning projects that are rooted in Jupyter notebooks. While not restricted to Oracle data sources, the new OCI (Oracle Cloud Infrastructure) Data Science service does come with connectors to the Autonomous Data Warehouse and cloud object storage. It will leverage the new OCI Data Catalog, which will be the metadata backbone of all Oracle analytic services (including the Oracle Analytics Cloud for BI and augmented self-service analytics) and bundled with them.
As a first release, Oracle still has some finishing up work to do on the new OCI Data Science service. Related services, such as the OCI Data Flow service (Oracle's answer to Amazon EMR and Kinesis) are not yet a unified experience, as you still have to boot them up as separate services. Another critical enhancement not yet in the first release is support for distributed training; initially, out of the box, OCI Data Science will only run on a single node. We expect that Oracle will add these features over the course of the year.
Meanwhile, back on the mother ship
The cloud has sparked the debate on the fit-for-purpose vs. multi-model database. Amazon has led the charge for bespoke databases, with 15 database platforms now in its portfolio. Meanwhile, Oracle, Microsoft, and SAP have placed their bets on the database as a jack of all trades (or at least, data types). Oracle is now amplifying that message by terming its database as a "converged database."
Oracle database has always supported multiple data types from text to XML, spatial, or Blobs, before it ever got to the cloud. But with some exceptions, those nonrelational data types were not first-class citizens. For instance, Oracle's JSON support was through representing it as a variable character string, which can contain a JSON document, but is not very efficient to query. Oracle is now stepping up its game in representing non-relational data types along with variants of relational data.
For instance, it is introducing a binary JSON data type that will improve query and update performance. This is not about compatibility with MongoDB, but about improving performance and supporting a wider range of data types. So, Oracle, like most other databases (excluding IBM Db2), do not support the BSON binary JSON format of MongoDB. There is compatibility if developers drop back to using standard, but less performant, character JSON data types.
Then again, given MongoDB's latest licensing changes, no database -- relational or NoSQL -- will be compatible with Mongo. The best that Amazon and others are doing is supporting MongoDB-compatible APIs providing access to core features, something we hope will be on Oracle's roadmap.
While Oracle binary JSON won't replace MongoDB, it will be useful for boundary cases where you have mixed relational and JSON data and want to take advantage of a mature SQL implementation.
Oracle is also introducing Blockchain tables that will still function as relational tables, but with the append-only and cryptographic support to make the tables immutable. And, it already includes a graph capability that will support RDF and property graphs, plus the ability to ingest IoT streaming data, and many built-in machine learning algorithms such as simple classification and complex neural networks.