How Microsoft re-envisioned the data warehouse with Azure Synapse Analytics

Rohan Kumar, corporate vice president for Azure Data, stands in front of a screen with the words “Microsoft Ignite”

About four years ago, the Microsoft Azure team began to notice a big problem troubling many of its customers. A mass migration to the cloud was in full swing, as enterprises signed up by the thousands to reap the benefits of flexible, largescale computing and data storage. But the next iteration of that tech revolution, in which companies would use their growing stores of data to get more tangible business benefits, had stalled.

Technology providers, including Microsoft, have built a variety of systems to collect, retrieve and analyze enormous troves of information that would uncover market trends and insights, paving the way toward a new era of improved customer service, innovation and efficiency.

But those systems were built independently by different engineering teams and sold as individual products and services. They weren’t designed to connect with one another, and customers would have to learn how to operate them separately, wasting time, money and precious IT talent.

“Instead of trying to add more features to each of our services, we decided to take a step back and figure out how to bring their core capabilities together to make it easy for customers to collect and analyze all of their increasingly diverse data, to break down data silos and work together more collaboratively,” said Raghu Ramakrishnan, Microsoft’s chief technology officer for data.

At its Ignite conference this week in Orlando, Florida, Microsoft announced the end result of a yearslong effort to address the problem: Azure Synapse Analytics, a new service that merges the capabilities of Azure SQL Data Warehouse with new enhancements such as on-demand query as a service.

Microsoft said this new offering will help customers put their data to work much more quickly, productively and securely by pulling together insights from all data sources, data warehouses and big data analytics systems. And, the company said, with deeper integration between Power BI and Azure Machine Learning, Azure Synapse Analytics can reduce the time required to process and share that data, speeding up the insights that businesses can glean.

What’s more, it will allow many more businesses to take advantage of game-changing technologies like data analytics and artificial intelligence, which are helping scientists to better predict the weather, search engines to better understand people’s intent and workers to more easily handle mundane tasks.

This newest effort to break down data silos also builds on other Microsoft projects, such as the Open Data Initiative and Azure Data Share, which allows you to share data from multiple sources and even other organizations.

Microsoft said Azure Synapse Analytics is also designed to support the increasingly popular DevOps strategy, in which development and operations staff collaborate more closely to create and implement services that work better throughout their lifecycles.

 

YouTube Video

A learning process

Azure Synapse Analytics is the result of a lot of work, and a little trial and error.

At first, Ramakrishnan said, the team developed highlevel guidelines showing customers how to glue the systems together themselves. But they quickly realized that was too much to ask.

“That required a lot of expertise in the nitty gritty of our platforms,” Ramakrishnan said. Customers made it overwhelmingly clear that we needed to do better.”

So, the company went back to the drawing board and spent an additional two years revamping the heart of its data business, Azure SQL Data Warehouse, which lets customers build, test, deploy and manage applications and services in the cloud.

A breakthrough came when the company realized that customers need to analyze all their data in a single service, without having to copy terabytes of information across various systems to use different analytic capabilities – as has traditionally been the case with data warehouses and data lakes.

With the new offering, customers can use their data analytics engine of choice, such as Apache Spark or SQL, on all their data. That’s true whether it’s structured data, such as rows of numbers on spreadsheets, or unstructured data, such as a collection of social media posts.

This project was risky. It involved deep technical surgery: completely rewriting the guts of the SQL query processing engine to optimize it for the cloud and make it capable of instantly handling big bursts of work as well as very large and diverse datasets.

It also required unprecedented integration among several teams within Microsoft, some of whom would have to make hard choices. Established plans had to be scrapped. Resources earmarked for new features would be redirected to help make the entire system work better.

“In the beginning, the conversations were often heated. But as we got into the flow of it, they became easier. We began to come together,” Ramakrishnan said.

Microsoft also had to make sure that the product would work for any company, regardless of employees’ technical expertise.

“Most companies can’t afford to hire teams of 20 people to drive data projects and wire together multiple systems. There aren’t even enough skilled people out there to do all that work,” said Daniel Yu, director of product marketing for Azure Data and Artificial Intelligence.

Making it easy for customers

Customers can bring together various sources of data into a single feed with Azure Synapse Analytics Studio, a console – or single pane of glass that will allow a business professional with minimal technical expertise to locate and collect data from multiple sources like sales, supply chain, finance and product development. They can then choose how and where to store that data, and they can use it to create reports through Microsoft’s popular Power BI analytics service.

In a matter of hours, Azure Synapse will deliver useful business insights that used to take days or even weeks and months, said Rohan Kumar, corporate vice president for Azure Data.

“Let’s say an executive wants a detailed report on sales performance in the eastern U.S. over the last six months,” Kumar said. Today, a data engineer has to do a lot of work to find where that data is stored and write a lot of brittle code to tie various services together. They might even have to bring in a systems integrator partner. With Azure Synapse, there’s no code required. It’s a much more intuitive experience.”

The complexity of the technical problems Azure Synapse addressed would be hard to overstate. Microsoft had to meld multiple independent components into one coherent form factor, while giving a wide range of people – from data scientists to line of business owners – their preferred tools for accessing and using data.


With Azure Synapse, there’s no code required. It’s a much more intuitive experience.”

~ Rohan Kumar, corporate vice president for Azure Data


That includes products like SQL Server, the open source programming interface Apache Spark, Azure Data Factory and Azure Data Studio, as well as notebook interfaces preferred by many data professionals to clean and model data.

“Getting all those capabilities to come together fluidly, making it run faster, simpler, eliminating overlapping processes – there was some scary good stuff getting done,” Ramakrishnan said.

The result is a data analytics system that will be as easy to use as a modern mobile phone. Just as the smartphone replaced several devices by making all of their core capabilities intuitively accessible in a single device, the Azure Synapse “smartphone for data” now allows a data engineer to build an entire end-to-end data pipeline in one place. It also enables data scientists and analysts to look at the underlying data in ways that are natural to them.

And just as the phone has driven waves of collaboration and business innovation, Azure Synapse will free up individuals and companies to introduce new products and services as quickly as they can dream them up, Microsoft said.

“If we can help different people view data through a lens that is natural to them, while it’s also visible to others in ways natural to them, then we will transform the way companies work,” Ramakrishnan said. That’s how we should measure our success.

Top photo: Rohan Kumar, corporate vice president for Azure Data, says Azure Synapse will deliver useful business insights that used to take days or even weeks and months. Photo by Scott Eklund/Red Box Pictures.

Related: