Microsoft has announced the general availability of SQL Server 2016 – which it says is the world’s fastest and most price-performant database for HTAP (Hybrid Transactional and Analytical Processing) with updateable, in-memory columnstores and advanced analytics through deep integration with R Services.
Software applications can now deploy sophisticated analytics and machine learning models in the database resulting in 100x or more speedup in time to insight, compared to deployments of such models outside of the database.
“The integration of advanced analytics into a transactional database is revolutionary,” explains Joseph Sirosh, corporate vice president of Microsoft’s Data Group, in a new blog post. “Today a majority of advanced analytic applications use a primitive approach of moving data from databases into the application tier to derive intelligence. This approach incurs high latency because of data movement, doesn’t scale as data volumes grow and burdens the application tier with the task of managing and maintaining analytical models. And deep analytics on real-time transactions are next to impossible without a lot of heavy lifting.”
According to Sirosh, SQL Server 2016 simplifies analytics in the way databases simplified enterprise data management, by moving analytics close to where the data is managed instead of the other way around. It introduces a new paradigm where all joins, aggregations and machine learning are performed securely within the database itself without moving the data out, thereby enabling analytics on real-time transactions with great speed and parallelism. As a result, analytical applications can now be far simpler and need only query the database for analytic results. Updating machine learning models, deploying new models, and monitoring their performance can now be done in the database without recompiling and redeploying applications. Furthermore, the database can serve as a central server for the enterprise’s analytical models and multiple intelligent applications can leverage the same models. It is a profound simplification in how mission critical intelligent applications can be built and managed in the enterprise.
“A good example of how our customers are benefiting from the new model comes from PROS Holdings, a revenue and profit realisation company that helps B2B and B2C customers achieve their business goals through data science,” Sirosh says. “PROS Holdings uses SQL Server 2016’s superior performance and built-in R Service to deliver advanced analytics more than 100x faster than before, resulting in higher profits for their customers.”
Royce Kallesen, senior director of science and research at PROS, says: “Microsoft R’s parallelisation and enhanced memory management on the server integrated with SQL Server provides dramatically faster results on a common platform with built-in security.”
SQL Server 2016 comes with several features and tools to support cross-platform analytics. “Polybase allows you to run queries on external data in Hadoop or Azure blob storage,” says Sirosh. “It can push computation to Hadoop where appropriate, so that your analytical application can join and integrate data from big data stores with the data in the relational store. Microsoft R Services, which is integrated with SQL Server also runs on multiple Hadoop distributions and is also integrated with Azure HDInsight + Spark, enabling both choice and standardisation in developing analytics code. And finally, R Tools for Visual Studio allows the ease of use of the modern Visual Studio IDE for developing analytical code in R.”
Sirosh goes on to explain SQL Server 2016’s ground-breaking performance optimisations and efficiencies, leading to new levels of performance and scale. For example, Microsoft recently collaborated with Intel to demonstrate stunning performance on a massive 100TB data warehouse using just a single server with four Intel Xeon E7 processors and SQL Server 2016. The system was able to load a complex schema derived from TPC-H at 1.6TB/hour, and it took just 5.3 seconds to run a complex query (the minimum cost supplier query) on the entire 100TB database. The system also demonstrated incredible concurrent query performance, where running all queries concurrently took less time than running them back-to-back, as illustrated in the chart below. This is a feat of scale, performance and efficiency that no other database has achieved to date.
Read more about the new release here.
Share this story