Startup Spotlight: Snowflake Computing's Elastic Data Warehouse
Created to redesign the enterprise-class data warehouse for a cloud-based world, Snowflake Computing's clients include Accordant Media, Adobe and DoubleDown.
It's been a busy year for Snowflake Computing, a data warehouse provider that officially slipped out of stealth mode last October.
The cloud-based company placed first in the Strata + Hadoop World 2015 Startup Showcase for its Snowflake Elastic Data Warehouse, which became publicly available in June. The startup also ranked among Gartner's 2015 Cool Vendors in DBMS and secured $45 million in its Series C funding this year, bringing its total funding to $71 million.
Yet despite the young company's accomplishments, its current CEO continues to garner the most buzz for Snowflake.
From Tech Titans to Startup
Snowflake's Bob Muglia isn't a typical startup CEO, even by Silicon Valley standards. Muglia was one of four presidents who reported directly to Microsoft CEO Steve Ballmer. He spent 23 years at Microsoft, including his tenure as president of its $17 billion-a-year Server and Tools division. Tech pundits questioned the company's future in 2011 when Ballmer released an email simultaneously announcing Muglia's departure and Ballmer's decision "to put new leadership in place for STB."
"I'm surprised Ballmer is considering a new president for the division," wrote Joe Wilcox of Beta News. Even Ballmer recognized there were few $15 billion businesses in the software industry, and Servers and Tools represented one of the three owned by Microsoft, Wilcox noted. His conclusion was less than glowing, however.
"From one perspective, Muglia is a dinosaur. He led the Server and Tools Business during the enterprise era of networked data. But Microsoft is looking beyond the corporate server market to cloud-connected data centers," Wilcox wrote.
Muglia had no intentions of being a dinosaur, though, and he would soon prove it. After a brief stint with Juniper Networks, he again made tech headlines when, in 2014, he became CEO of an unknown cloud-based startup.
"What I recognized is that all these companies have very large existing legacy businesses that were built in the pre-cloud era and were not going to grow on a go-forward basis. In fact, the growth opportunities that are happening are all being created by much smaller companies," Muglia said. "Rather than work on legacy products for the rest of my career, I wanted to build something."
Muglia wanted a cloud-based company targeting the enterprise space with something "incredibly innovative and incredibly different," he said. Former Juniper colleague Stefan Dyckerhoff now worked at Sutter Hill Ventures and had shepherded Snowflake through its first round of funding.
Data Warehouse Superstars
Snowflake was the brainchild of three established stars from the data warehouse world: Benoit Dageville and Thierry Cruanes, who previously worked together at Oracle, and Marcin Zukowski, who led development on the high-performance query processing engine X100 (later renamed Vectorwise). Dageville had worked as lead architect for parallel execution in Oracle RAC and a key architect in the SQL Manageability group. Cruanes specialized in query optimization and parallel execution at Oracle, but he had also worked on data mining with IBM's European Center of Applied Mathematics. The men wanted to build a cloud company that would offer data warehousing as a service.
All three hold Ph.Ds in computer science so they began their venture by examining the source codes of existing data warehouse architectures, including Hadoop running MapReduce. They soon discovered the code bases included an obsolete assumption: that compute power must be tightly coupled with the data.
"The most important thing here is that those architectures are all built under the premise -- which had been widely established in the industry -- that in order to analyze large amounts of data, you need to move the compute to the data," Muglia said.
What Makes Snowflake Special
Three relatively recent advances made that assumption obsolete, Muglia explained. First, BLOB storage was now both highly durable and scalable. Second, the cloud allowed companies to spin up any size clusters of compute power on-demand. Third, data can now be moved quickly over 10 gigabit Ethernet networks. Technology had evolved, but the DNA-level assumptions of data warehousing were stuck in the Mesozoic Era.
"In order to build a system that had the characteristics that they wanted to build, they couldn't leverage any of those source codes," he said. "They needed to build something from scratch, and the results are pretty extraordinary."
By de-coupling storage from compute, Snowflake Elastic Data Warehouse can scale linearly and compute faster. This makes it particularly useful for Big Data sets such as unstructured, machine-generated data (think Internet of Things) and transactional data.
The decoupling of data from computing also allowed the company to offer something unique: It can scale the number of users working against the data while maintaining only one version of the data, Muglia said.
"We can elastically scale the number of users that are working against that data, because you can have multiple clusters working against the data at the same time with the data stored in only one place," he said. "Nobody else can do that." Essentially you can put an infinite number of users on Snowflake.
"In fact, because we maintain transactional consistency across that data, writers don't block readers. So you can be writing data and reading it at the same time with multiple users working," he said.
That feature is likely to appeal to large enterprises, who must balance their hunger for data-driven decisions against compliance concerns. But will it be enough in a $12 billion industry that includes technology titans such as Amazon's Redshift?
Muglia says yes, citing the company's affordability and focus on enterprise-class requirements such as turnkey security features and certifications for its data warehouse. (The company plans to have HIPAA certification by next year.)
"There's this huge transition happening as people began to evaluate and move from on-premise deployments of things like Oracle, Teradata and Netezza, and began to focus on cloud-based solutions," he said. "We're really focused on positioning Snowflake as the leader for cloud data analysis and data warehousing. We also solve the problems of working with the Big Data, the semi-structured data, much better than anything else on the market – much, much better than Hadoop alternatives."
Fast Facts about Snowflake
Founders: Benoit Dageville, Thierry Cruanes and Marcin Zukowski
HQ: San Mateo, Calif.
Product: The Snowflake Elastic Data Warehouse is a data warehouse as a service, built from the cloud up
Customers: Accordant Media, Adobe, DoubleDown, KIXEYE, WhiteOps and The Orchard
Funding: $71 million, with investors including Sutter Hill Ventures, Red Point, Altimeter Capital and Wing VC
Loraine Lawson is a freelance writer specializing in technology and business issues, including integration, healthcare IT, cloud and Big Data.