A raft of expansions across Microsoft's Azure cloud database tier adds capabilities that target high-end workloads and shops that prefer open source options.
SEATTLE -- Microsoft's latest string of Azure cloud database technologies looks to address a multitude of developers' needs that range from high-end workloads to edge applications.
At the Microsoft Build 2019 Developer Conference here this week, Microsoft disclosed that Azure SQL Hyperscale is now generally available. The database raises the 4 TB limit on Azure SQL database sizes up to 100 TB.
previewed in October, the Hyperscale tier will assist shops that previously had to split large Azure SQL workloads across multiple databases. Admins can now manage actions such as database recovery within a single large instance.
Microsoft also a Hyperscale option for PostgreSQL into preview. It is based on the January acquisition of Citus Data and added as an extension to PostgreSQL, not a fork of the core codebase, which means that customers can keep in step with updates.
The PostgreSQL expansion is Microsoft's latest push into open source database options for Azure. "We are doing great things with proprietary databases, but Azure is a platform for every developer," said Rohan Kumar, corporate vice president of Azure data platform, at Build.
Azure SQL Database serverless is also in preview. The fully managed service is best for intermittent workloads on a single database, for scenarios such as development and test, gaming and e-commerce, Kumar said.
SQL Database serverless is billed on a per-second basis, which saves customers money over less granular models, Kumar added.
Chemical giant BASF uses Cosmos DB for a project that draws data from 20 million sensors spread across hundreds of sites around the world to compile data on metrics such as temperature, pressure and flow rates, said Robert Pack, a research scientist who joined Kumar onstage.
"We're not looking to optimize production within the scope of a single plant, but rather optimize production across multiple sites and value chains," Pack said. "Exploring this very high-dimensional space and arriving at actual insights for our experts is a tremendous challenge."
BASF has used Cosmos DB to build a graph representation of its footprint that provides thermo-physical and systems-level contexts for all the signals it collects. This provides a basis for analytics it runs with the help of Azure Data Explorer, according to Pack.
"We're now able to analyze decades of high- time-series data from multiple instances of a process, distributed across the globe, to build models that not only are valid for one specific plant, but a technology as a whole," he said.
Microsoft also disclosed a preview of Azure SQL Database Edge, an optimized version of the database geared to run on edge devices for IoT scenarios.
Choosing the right database begins with the right discussions
The cohort of Azure cloud database updates at Microsoft Build are on the whole interesting but should be taken in the right context, said Jarrad Edwards, CTO of Roam Creative, a digital services consultancy with offices in Sydney, Berlin and Auckland, New Zealand.
I'm not a believer of just taking any database that looks like it will fit and pushing it into a project. It's got to be the right database that has the right benefits for the end use case.
Jarrad EdwardsCTO, Roam Creative
"I'm not a believer of just taking any database that looks like it will fit and pushing it into a project," Edwards said in an interview after Kumar's session. "It's got to be the right database that has the right benefits for the end use case."
Customers shouldn't be dazzled by new capabilities and neglect to have a solid, broad-based discussion with various constituents on the project's ultimate goal, he said.
"Technology is only one side of it," Edwards said. "Myself, being a CTO, I would sit there and debate with a CIO, to talk about what their needs are so we can make sure we get a database that benefits everybody in the picture."
Edwards recalled an application project he worked on several years ago.
"We took a database that was rather read-heavy -- you could pull a lot of information out quickly in succession -- but as our application grew, we found it was actually very write-heavy," he said. "If we had spent more time upfront and actually discussed it as a whole ... we would have gotten to a solution so much quicker."
Azure cloud databases' all-things-to-everyone approach not unique
Of course, Microsoft isn't the only vendor that pushes a phalanx of database options, particularly open source ones. At the same time, cloud providers also steer customers to their own proprietary database services, said Doug Henschen, an analyst at Constellation Research. Microsoft, for example, touts Cosmos as being "wire-protocol compatible" with open source options Cassandra and MongoDB, while AWS positions Aurora as a compatible replacement for MySQL and PostgreSQL.
Compatibility in that sense has significant limits, according to Henschen.
"Customers should keep in mind that proprietary, single-cloud services are not portable to other clouds or to conventional on-premises deployment," he said.
It's also not a simple matter to migrate a deployment back to an open source database, if and when you want to move it back on premises or to another cloud.
"The point is avoiding placing all bets on one cloud and being very choosy about using single-cloud services that might diminish deployment flexibility," Henschen said.
