SQL Server for Linux makes a lot of sense to me. From the technical perspective, I'm surprised that it's not been done before. SQL Server still has the fundamental architectural design it inherited from Sybase, which allows it to be adapted quickly to new operating systems.
The Least Recently Used (LRU) buffering system goes back a long way, but essentially it was done to make the system extremely portable, because it abstracts away the file level entirely. Of course, a lot of dependencies that specify files have been put back in since those far-off Sybase days, as we all know to our cost when deploying to different hardware configurations in servers. After many years being able to ignore portability in the product, a portable version of SQL Server is once more a goal worth pursuing, even where SQL Azure is a viable alternative. It is hardly surprising, therefore, that from the first announcement in March, a version of SQL Server is now being previewed on Ubuntu and Red-Hat with 'core relational database capabilities'.
I've past experience, long before there was the option of SQL Azure, of having to introduce Windows-based SQL Server to an entirely Linux-based infrastructure. Although it got us a good fast database with all the features we needed, the maintenance overhead was horrible. You need the expertise in Ops to maintain a production Windows server around the clock, and you need a whole set of maintenance systems especially for that single Windows system: stuff like monitoring, intrusion detection and alerting. It was an experience that still gives me the occasional flashback.
Having a native Linux SQL Server is enough to make Ops people smile, but more is due. Microsoft have already demonstrated a SQL Server on Linux inside a Docker container. Once this gets to production standard, it is going to give Ops staff a versatile way of providing the right amount of database capacity, in the right place, at the right time.
There is another glint in Microsoft's eye, and that is making it easier for Oracle users to jump ship to SQL Server. I get the feeling from friends working with Oracle Linux that having a native Linux SQL Server would give them a much better economic case for doing the move. Given the number of Oracle Linux users, this is surely a demographic worth pursuing.
I'm assuming that this work will go hand-in-hand with all the other Microsoft initiatives towards a more portable platform, such as C# and ASP.NET on Linux. It has been a major initiative for some time and it is likely to be good news for those of us who are employed as SQL Server experts. The Microsoft data platform is now getting plenty of room to grow.