This article is more than 1 year old
The Register calls for aid, and Microsoft's Rohan Kumar will answer... our questions about SQL Edge and Azure Synapse
Slimming down SQL and living on the Edge
Build There was SQL Edge and Azure Synapse news at Microsoft's reimagined Build gathering this week, so The Register had a chat with corporate vice president Rohan Kumar about the company's database ambitions.
Having lurked in limited preview for a while, Azure SQL Edge has been pushed out to a broader audience as the platform creeps toward general availability. Running on x64 or ARM, with Windows or Linux doing OS duty, Azure SQL Edge is all about shunting the smarts of its bigger SQL Server brother to edge devices for more local processing and storage rather than maintain a constant, and potentially laggardly and expensive, connection to the mothership.
Kumar cited AI training as an example use case for the technology. "[We] see a lot of customers using our big data analytics in the cloud to train and run their machine learning models," he said, "and once they come up with a model that they believe is meeting the requirements, then they mass deploy onto the SQL Edge devices."
Those devices can then run in environments that are online, occasionally connected or fully offline.
"A lot of decisions can then be made very close to where the data originates," he added.
Kumar told us that the content of the wider public preview was pretty much what would go on to hit general availability, although the gang still had a way to go in order to reduce the current 500MB footprint down to 300MB. The limited capacity of Edge devices, such as the popular Raspberry Pi, makes the bloat loss essential.
"This is not the SQL Server of 15 or 20 years ago," laughed Kumar. "It runs everywhere." Indeed, the first SQL Server this writer used was version 4.21 and it ran happily enough under NT with a massive 64 megabytes of memory. Still, it is heartening to see the footprint being reduced while other products continue to pile on the pounds.
Azure Synapse Link
Also hitting public preview, and some way from the grungy edge shenanigans of SQL Edge, was Azure Synapse Link, a "cloud native implementation of hybrid transactional analytical processing (HTAP)", according to Microsoft.
Kumar had previously shown off Azure Synapse, a revved-up and rebranded of Azure SQL Data Warehouse (or "evolution", as Microsoft would have it), and its analytical capabilities at last year's Ignite event.
Synapse also gained some new toys in Public Preview at Build (something Kumar described as a "milestone"). The analytics service added new Synapse SQL features and, inevitably, built-in Power BI authoring. The widening of the preview also means that more users can check out if the team's performance and ease of use claims cut the mustard.
However, getting data from operational systems into analytical services has traditionally been a bit of pain. The new Link functionality allows customers to hit that real-time transactional data in Azure without adding a burden to operational systems.
There are, of course, gotchas.
It is undoubtedly neat if you're in the Microsoft ecosystem, but right now Azure Synapse Link only supports Azure CosmosDB in the preview. Other Azure databases such as PostgreSQL, MySQL or even Azure SQL are still in the "coming soon" bucket, which is a little disappointing.
Kumar assured us that the gang were "actively working on both SQL and Postgres", but that "there is a certain set of priorities that every team works through" and CosmosDB came out ahead.
On-prem or not on-prem
A frequent complaint levelled at Microsoft is that much of the new stuff depends on the Azure cloud while many customers still prefer to keep their data close at hand. Azure Stack is an oft-touted solution to the problem, but Kumar pointed to Microsoft's crack at multi-cloud, Azure Arc, for an indication of where Synapse might go next.
"That could be the next step in the journey," he said, "if a customer is willing to maintain connectivity with Azure, where we're able to get a certain amount of telemetry, then progressively we can add more and more capabilities."
The cloud-versus-on-premises argument continues to rage concerning SQL Server 2019. Could that be the last major on-premises release of the former flagship? "No, absolutely not," stated Kumar. However, he added: "Here's the thing: if you look at the kind of investments we are making in SQL Server, even on-prem, it's essentially not just to support the customers, but to prepare them for the transition to the cloud."
That transition continues to rumble on as Microsoft attempts to persuade those still on the elderly 2008 and 2008 R2 platforms to move to something more modern (faced with the IT curse that those boxes simply work, and most admins know that if something works, it is best not to go fiddling with it).
And the on-premises versus cloud balance? At present it stands at around half and half. There are an awful lot of SQL Servers still alive and well in data centres around the world. ®