Launched in 2012 after 2 years in stealth mode, cloud data warehousing company Snowflake Computing kicked off 2018 by announcing $263 million in additional growth funding. Since its founding 6 years ago, Snowflake has raised a total of $473 million in growth funding with the mission to combine the power of data warehousing, the flexibility of big data platforms, and the elasticity of the cloud. Recently, Christian Kleinerman, the company’s VP of product, reflected on how enterprise data architecture is changing and the growing role of cloud in analytics.
Enterprise data architecture has changed greatly over the last decade. As a company that provides a data warehouse built for the cloud, what is Snowflake’s point of view on key aspects to helping organizations to get the most from their data?
Christian Kleinerman: Today, there is a focus on building architectures that are insight-centric or insight generation-centric. This means that the emphasis is going from infrastructure, data warehouses, and technology to a focus on what is needed to gain insights and then tools are being adapted around that primary issue. Organizations don’t want to focus on managing networks or servers; they want to focus on capabilities to answer questions. Essential elements of this are the notion of right-sizing capacity. Not too long ago, customers ended up having to choose a fixed configuration based on data volume, or compute and storage ratios. Snowflake has been very focused on right-sizing the offering. This means getting the compute demand that you need at the time that you need it and only paying for what you use. One of the benefits of the cloud is this notion of virtual limitless resources. A second important aspect is technology that does some self-tuning to allow users to focus on their business questions and insights. And the third is around flexibility of priorities.
What does Snowflake see as the key advantage of a cloud-enabled data warehouse?
CK: Radically new at the beginning, it has been 6 years since Snowflake was founded to leverage all of the insights and flexibility of the cloud. For example, being able to have unlimited storage. The separation between compute and storage is a massive difference from traditional data warehouses. In the appliance market today you have to say how many terabytes you want up-front, how many servers, how much memory up-front as opposed to getting the capacity as needed and paying for what is used. Snowflake has leaned into all of those trends that the cloud provides as benefits.
Do many customers have mixed hybrid physical/cloud infrastructure or multi-cloud deployments and what is the thinking behind those approaches?
CK: Different companies are at different stages of the cloud adoption cycle. Some are primarily on-prem, while there are newer companies that typically are entirely in the cloud. As recently as 5 years ago, people at Fortune 500 companies were saying, “Do not mention the C word. I am not interested in it.” But everyone at this point is willing to test it at least, or go all-in on the cloud.
Enterprises have also learned about vendor lock-in in the on-premise world and for that reason many companies have standard policies of using two database companies, two hardware providers, and the same idea is transferring to the cloud where some companies are saying they would not want to be completely dependent on Amazon or Microsoft or Google Cloud. They want to have a second or third cloud provider either for different workloads, disaster recovery, or load balancing. It is a way to have insurance and, in some ways, to have leverage over cloud providers.
Outside of the major cloud providers—Google, Amazon, Microsoft, and maybe Alibaba—you see a number of companies that are starting to emerge and provide cross-cloud or multiple cloud services. And while Snowflake is only on Amazon now, it is our plan to go to other clouds because we want to provide choice for our customers.
So, you will be expanding to other cloud platforms beyond AWS?
CK: Directionally, that is where we are going. We want to meet customers where they are.
Everyone wants to be data–centric. Are there any issues you see as companies try to get more value from data?
CK: Yes, one is know-how, the other is tools, and the third is data availability.
Know-how is sometimes underestimated. It is not just collecting data. It includes an understanding of metrics, what makes good analysis, and how to remove biases from analysis. The second is the shift from what tools let you do, to selecting tools to accommodate the decisions you are making. If you think you should be looking at customer behavior over the last 30 days or the last 3 years, the technology is there to enable that level of very large-scale queries, analyses, and insights, and if you want to do predictive analysis or forecasting, the technology is there. A key tenet of Snowflake’s service is that we want to make things as simple as possible with auto-tuning so that the barrier to entry to make good data-driven corporate cultures is as low as possible.
And, data availability?
CK: Frequently, some of the data that you need in order to obtain real insights lives outside of your group or company. For example, if you are the CEO of a large multinational corporation, the sales from each country or the growth rate in each country is not as interesting if you cannot contrast it with the GDP growth for each of the countries. Making that data available from different data owners into companies is a priority for us. The spirit behind it is that as data is shared between departments, companies, or supply chain members, there is more value to be unlocked from it. Through data sharing at Snowflake, we are trying to help companies get the right data to make sure that they get the most out of their environments.
There have been some high-profile data breaches recently and also new data handling regulations such as GDPR are coming into play. How significant are those concerns in terms of the cloud and what kind of assurances are companies seeking?
CK: Security and privacy is a high priority for every customer, but the most interesting thing that I have noticed is that while 5-8 years ago, security was the reason to not go to the cloud, today security is one of the most important reasons to move to the cloud.
If you look at what happened with many of the breaches, it was a problem in the processes, people, and IT departments as opposed to an issue with not having the right technology. In one high-profile breach, there was vulnerability in a piece of software and the IT department was supposed to patch it but they didn’t find all of the instances and they got hacked. With the cloud, you are leaving all that to the cloud vendors which obviously is a very big part of the value proposition. With Snowflake, customers can leverage the expertise and, in some ways, economies of scale because we or Amazon are handling it on behalf of many.
Are there any trends that you expect to emerge or grow stronger over the next 3 years in terms of what organizations are focused on?
CK: I don’t think everything is going to move overnight to the cloud but every company is identifying a use case where cloud is a priority. BI tools in particular are playing a role as an abstraction or a model on top of both on-prem and cloud data and making it very easy for customers.
The second trend is around more continuous data injection. Many companies still use daily batch or weekly processes but increasingly you see people talking about continuous data ingestion or streaming technologies because the time to insight matters.
And then the other piece is that if you can look past the hype of machine learning and AI, predictive analytics, which has been emerging for several years but has been difficult to achieve, has evolved to the point where every company I talk to is trying to figure out how to leverage it. That is a big trend that is becoming mainstream.
Is there anything that concerns you as you look at data management and analytics?
CK: There is a shortage of data scientists and people who can do data science. Every company will benefit from a data-driven culture and data-driven decision making processes and it will be interesting to see how the market evolves. The market has to make data management processes easier to use so that data scientists can focus on more advanced use cases but other tasks can be accomplished without needing a Ph.D. in statistics.