Tech is supposed to be a young person’s game, which makes the perspective of someone like Snowflake Computing CEO Bob Muglia even that much more valuable.
Muglia, who ran Microsoft’s Server and Tools Division for five years as the capstone to a 23-year-career at the software giant, now runs Snowflake Computing, a 300-person data warehousing startup that thinks it has developed the best cloud-native product for analytical databases. Cloud databases are turning into a very important segment of the enterprise tech market, as Oracle attempts to hold on to its legacy customers while cloud providers and others like Snowflake try to pull them in a new direction.
I caught up with Muglia on the sidelines of Amazon Web Services’ re:Invent 2017 conference in Las Vegas last week to talk about shifting database market trends, the pecking order in infrastructure cloud computing, and how Seattle has changed since he first arrived in 1988. A lightly edited version of our talk follows below.
GeekWire: I’m interested in hearing about your cloud provider strategizes, and hope to get an idea of where databases are going over the next five years.
Bob Muglia: We’re five and a half years old as a company. Our founders recognized you could build a cloud, a data warehouse, a relational database very differently using cloud-based technology than you can in the traditional on-premises world. They recognized this five years ahead of the industry. I think people are reading our white papers, and then figuring out, “OK, this is the way to do it.” We may have some people building something similar now, but our team’s been way, way, way ahead of that.
GeekWire: What are those key differences that you would identify?
Bob Muglia: The attributes that eventually customers see are basically, we are a SQL relational database, an analytic relational database, likened to a Netezza, Teradata, or Vertica, something like that. We’re not an OLTP database. We don’t replace Oracle or MySQL, but we can replace and have replaced all of those other products. We do replace Oracle and SQL Server for things that are more analytic in their nature. While we look like, and are in fact a full transactional SQL database, we don’t have any limits that these other systems have.
GeekWire: Such as?
Bob Muglia: (The) amount of data you can store, how many queries you can run concurrently, those are the two big ones. How much data can be processed, and how many queries can we run concurrently.
The difference is, all those other systems are built using architectures that go back 30 years, literally 30 years, going back to the mid-1980s. Most of them are built on an architecture called shared nothing, where you have a cluster of machines with the data coupled to those nodes, and that cluster and the computers within it cooperate to run the queries. In those systems, that cluster defines the unit of scalability. That cluster consists of some number of physical machines. You can process more data by adding more nodes to the cluster, but you can’t really run more queries simultaneously by adding nodes. That’s gated by how much memory there is in those systems.
All of these systems have some limit in terms of how many queries they can run concurrently, five, ten, twenty, maybe a really scalable system can run 30 to 50 queries concurrently, but you hit that limit and you are done. You can’t run any more queries. We don’t have those limits because we have the ability to have multiple clusters running against the same data at the same time with full transactional support.
Nobody’s ever been able to do that. We’re the only product on the market that does that, the only product on the market that can do that. What that means is that there is no limit to the number of queries you can run with full transactional consistency associated with it. This is a very big deal because when we talk to customers what happens is, with these other systems they are successful. The systems are mature, and they work, and they’re good products. They start with a good experience, but then as they put more load on it, and they put more users on it, they hit these limits. Then their users complain because their stuff runs really slowly.
There’s very little they can do to fix that problem. The only way to fix it is to introduce multiple clusters, multiple machines that are each independent. You have to manage that environment, which is … now your data’s all over the place. It’s not all in one database. You have all these different systems, and you have to send some users over here, and some users over there. It’s just a mess. With Snowflake, none of those problems exist. You can have all your data accessible to all your users all the time.
GeekWire: Are there specific applications that take advantage of that?
Bob Muglia: Sure, there’s a couple. Anybody who is offering using their data warehouse as a part of a service offering, is going to hit that limit. If you’re a SaaS provider, you’re going to hit that limit. If you become successful, you will hit limits guaranteed, guaranteed. But any large enterprise will hit it. If you can only run five or 10 queries simultaneously, that’s a pretty small number.
We literally hear customers that it takes three days to close the books at the end of every month. That’s a very predictable (event) and they don’t like that, and they don’t like that very much. We had one customer that had to do their consolidated billing at the end of every week, and it literally took them the entire weekend to run this. Assuming everything went well, it took them the entire weekend to run this job, and it had to be done on Monday morning. In literally the best cases it got done like Sunday night at 10 PM. If anything goes wrong, they’re in trouble.
With Snowflake, they don’t have that problem because now they literally every Friday afternoon they throw multiple warehouses against it and they get the whole thing done in just a few hours.
GeekWire: So, you’re primarily on AWS?
Bob Muglia: We’re only on Amazon at this point.
I mean, look; we are a cloud data warehouse. We were built and born on Amazon. Amazon is great, they are a fabulous partner. We do have customers that ask for other clouds. We are a customer-centric company. We’re a values-based company. Our first value is we put our customers first. And there are customers that want other clouds so we’re certain…
GeekWire: Why do they want other clouds?
Bob Muglia: Couple of reasons. First of all, there are some industries that are somewhat averse to Amazon. For example, the retail industry.
GeekWire: Is that growing (into more of a thing)?
Bob Muglia: It’s pretty real. If anything, we’d say it’s simply this: in talking to quite a few companies in the retail industry, all of them noticed the Whole Foods acquisition.
I think that that galvanized the retail industry, and I think they have … obviously Amazon is a very significant competitor to them. And to be fair, none of these companies think that Amazon is looking at their data. I don’t think anyone is worried about that. And certainly, if their data is on Snowflake, Amazon couldn’t see their data because it’s all encrypted and Amazon would have no way of looking at it because they don’t have the keys.
But their bigger concern is simply, “Why should I spend my money with a competitor?” And that is pretty real. And it’s an understandable concern. The other thing I’d say is that customers have preferences in vendors. And you have a lot of customers for sure that have worked with Microsoft, as an example, for a long time, so that some reasons they do.
GeekWire: Do you see traction with Google, or IBM, or any of the other clouds?
Bob Muglia: We do see some, but we see a lot more, we hear more … I mean first of all let me be very, very clear. The vast majority of our customers want Amazon. But when we hear other clouds, I would say nine times out of ten, it’s Azure. And then IBM sometimes and Google sometimes.
I think, in terms of other clouds, and longterm viability of other clouds, I’m very clear that Azure has a longterm future. I think it’s likely that Google will be successful. And I’m not sure that there will be a whole lot after that. We’ll have to see if SoftLayer … how successful IBM is on a long-term basis.
And then the other sort of wild card in all this might be Oracle. And I think that Oracle has a captive customer base … ‘captured customer base’ might be a way to put it. And they certainly are doing some things economically to incent that customer base to go to Oracle. So I think you’ll see that. I think what we’ll see is essentially all of Oracle’s customers will have some adoption. I mean that a little overstating it, but I think we’ll see very broad adoption of Oracle Cloud, for something where Oracle is used. But I don’t see people who are not in that space right now moving to Oracle.
I sort of look at this thing fairly simply. The (global) IT market is roughly a bit over three trillion overall. About half of that is communications and the other half is everything else. So you’ve got 1.5 trillion in IT spend, basically.
It started with 0% cloud adoption. I think it’s realistic that in ten years half of that spend is going to be in the cloud … at least a third, 500 billion to three quarters (of a triillion). You look at this and Amazon is on their way to building another $100 billion business for themselves.
GeekWire: I can’t let you go without talking to you about Seattle a little bit. What do you make of the changes in the region over the last 15 to 20 years?
Bob Muglia: Well, you know, it’s funny. When I moved to Seattle 30 years ago. Literally it will be 30 years in January. When my wife and I moved to Seattle we were in the Bay Area and we consciously left the Bay Area because we felt it was too crowded and too expensive, in 1987. In 1987.
Bob Muglia: And when we moved … this is all a true story … we just thought it was very crowded and expensive. When we moved up to Seattle. I said “My greatest fear is that Seattle is 20 years behind the Bay Area.” And I actually think it’s probably 15.
And what we’ve seen is, as the Bay Area has grown to a point where it’s sort of, I don’t know how it grows too much more. As that sort of happened, I think that people have recognized that there needs to be other places where we have technology centers of excellence. Seattle has really emerged, more so than ever, as the very strong number two after the Bay Area.
As the CEO of a growing software services company that’s based in the San Francisco area, Seattle is very attractive because it’s the same time zone and it’s a two hour flight, and there are a lot of flights. And when people say, “Well, what about Vancouver?” It’s like well, Vancouver is great, but you have to cross the border. And there’s only a couple of flights a day. I mean there are flights, but they aren’t like Seattle.
GeekWire: Are you finding recruiting harder?
Bob Muglia: It is so much easier in the Seattle area than it is in the Bay Area for us. And it was my expectation that that would be true and it certainly is turning out to be. We can find really high quality people. It’s not cheaper. We never thought it would be less expensive. We pay people the same. Our pay rates are the same. Now they do get the benefit of the income tax difference between Seattle … Washington and California. So there’s an automatic ten-ish percent difference in comp associated with that.
So it’s actually easier for people to live on the same salary in Seattle than it is in the Bay Area. But we never thought it would be cheaper and it’s not cheaper.
For us, we have a very attractive value proposition to a potential candidate. We’re a growth company and we’re small but growing rapidly. We’re not an early stage startup, so the risk of coming to Snowflake is relatively low. It’s higher risk perhaps than going to Microsoft, but it’s very low risk overall.