It takes a certain amount of courage to compete with public-cloud giant Amazon Web Services — especially when your own product runs exclusively on AWS. But that isn’t fazing Bob Muglia, a 23-year Microsoft vet who in June 2014 stepped into his first C-suite job as CEO of data warehouse-maker Snowflake.
“We compete and we’re good partners at the same time,” Muglia said. “Don’t forget, AWS sells compute and storage time whether someone’s running Redshift (its own data warehouse) or Snowflake, so our success is their success, to some extent.”
Muglia, now 53, shared his candid views on that relationship, on Azure’s shortcomings and on what drove him to leave Microsoft during a one-on-one interview with GeekWire at AWS’s re:Invent user conference in Las Vegas.
Coopetition with Amazon Web Services
Snowflake, a San Mateo, Calif., startup founded in August 2012, makes a data warehouse that competes directly with AWS’s Redshift. Data warehouses keep copies of structured data (in rows and tables, such as relational databases) or, in some cases, semi-structured data (JSON files, logs, mobile-device output and Internet-of-Things sensor output) and analyze it to reveal unsuspected connections and patterns. They’re old technology, at least as offered by companies including Oracle, Neteeza, Teradata and Vertica, in Muglia’s view.
“Our approach is different from those and from Redshift, which is built using the architecture all the others also use, called ‘shared nothing,'” Muglia said. “They all involve a set of compute nodes closely coupled to data, working together as a cluster. But for any cluster, you will eventually run out of capacity, and adding new clusters requires creating new ones, copying and data and managing it. It’s very difficult for companies to do that.”
Snowflake, in contrast, scales easily, he said. It uses AWS’s S3 as its sole storage mechanism and has a three-layer architecture: storage on the bottom, compute in the middle and service (management) on top.
“When a Snowflake cluster runs out of capacity, we can instantiate another cluster right next to it, accessing the same data at the same time, and basically continue to scale,” Muglia said. Conventional data warehouses “can’t be turned into what we’ve created, which we call multi-cluster shared-data architecture.” Snowflake also allows running relational queries and updates on stored data, a feature missing from many open-source competitors, he said. And its ability to process semi-structured data is a feature most data warehouses omit.
Snowflake runs only an AWS, Muglia said, because “S3 is such a sophisticated service. Azure’s blob storage is not as sophisticated, so it would be a technical challenge to run on Azure. This is one of the areas where Amazon is just ahead.”
Muglia didn’t hesitate to bash Azure, in the process taking a swipe at himself.
“We obviously compete with the cloud vendors, for example with companies that were using (Microsoft’s) SQL Server (relational database) and are looking at Redshift, Azure, Snowflake,” he said. “When we have those (as competitors), we tend to do very well, because Azure’s products are very deficient, particularly Azure Data Warehouse. Redshift is a much better product, which I can tell you because (Azure Data Warehouse) is technology based on technology that was built by my team.”
AWS “has done a great job of supporting us,” Muglia said. “We’re very happy with them. They helped sponsor our recent tour of 10 cities. Their sales reps come out on our sales calls under a program Amazon has put in place. But we compete for customers. We talk about our benefits and they talk about theirs.”
The data warehouse market is worth fighting for. For conventional products alone, IDC values it at $14-$15 billion, “but the cloud data warehouse market will be much bigger, because the cloud will be larger and you can do a lot more in the cloud,” Muglia said. “All the legacy-technology (end-user) companies will move to the cloud eventually. ”
Microsoft ‘missed the shift in the industry’
For 23 years, through 2011, Muglia labored in the engineering arm of Microsoft, at various times building SQL Server, running developer operations, the server group and Office, then running Windows and the entire server and tools business. In the 2000s, the early days of open-source software, “I got to feeling there was a lot more we should be doing with the open-source ecosystem, but Steve (Ballmer, then CEO) was well-quoted as calling it cancer,” Muglia recalled. “Microsoft had a policy that was very protective of its intellectual property, and I always felt that was very detrimental, because we weren’t participating in the ecosystem.
“And Steve was, like, ‘No, we need to hang onto this, because Linux will be competing with Windows.’ It almost got to the point where we were making threats that people who were using open source needed to license it. And I was just, like, ‘This is broken.'”
Ballmer and Microsoft co-founder Bill Gates “were in sync on this,” Muglia said. “But the world changed and they hadn’t. They were operating in the business model of the 1980s, not the model of the 2000s. It’s true that by retaining that intellectual property, Microsoft has a $3-billion patent-royalty business right now. Whoop-de-doo. That’s not going to move the mark, yet they missed the shift in the industry because of it.”
Under current CEO Satya Nadella, Microsoft “is positioned to do much better, because one of the first things he did is reverse policies,” Muglia continued.
But by the time Nadella was appointed CEO, in February 2014, Muglia had left for Juniper Networks, where he ran the software group. Flat growth, budget cuts and quarter after quarter of layoffs proved discouraging, so he began searching for something new.
Avoiding the daily ‘BS’
“My time at Juniper helped me see that the transition to the cloud was going to be a major tidal-wave disruption and that the companies I call old technology — HP, Dell, Symantec, SAP, Cisco, Oracle — were going to struggle as a result,” he said. “I realized that most of the innovation was happening in smaller companies. And for the remainder of my career, I wanted to build something that matters. I was looking for a big market, a major disruption, something cloud focused. And I love databases. I heard about Snowflake through Sutter Hill, an early investor and builder of the company. When I heard what they’re doing technically, I recognized it could really work. So essentially, I was hired by the founders of Snowflake.”
What’s it like being a CEO for the first time?
“It’s fun!” Muglia said. “People say it’s a huge amount of work to lead a startup, and it’s true. But it’s rewarding because it makes a difference. At Microsoft or Juniper, a large percentage of what I did every day was BS — just stuff you had to do to make the business go forward, just overhead. There’s no overhead at Snowflake. Everything you do is focused on helping the team to grow, or improving the product, the sales process, the marketing.”
His biggest challenge, he said, is “scaling, scaling, scaling. In particular, hiring engineers is one of my top challenges.”
To that end, Snowflake is looking to open a Seattle office with 10-15 people initially, because “it’s a very strong market for systems talent, with the history of Microsoft and Amazon.” The search for real estate in Seattle is set to begin immediately, he said. “I hope to grow it to be a a material office over time.”
For now, Muglia is commuting between Seattle, where he and his wife own a Mercer Island house, and San Mateo, where they own a condo. “My driver’s license is issued by California and my wife’s is from Washington,” he said.