Our Chief Product Officer, Lenley Hensarling, recently appeared as a guest of Arman Eshraghi, host of the podcast “SaaS Scaled.” Esharghi, CEO and Founder of Qrvey, a provider of embedded analytics for SaaS applications, invited Lenley on his show to discuss the impacts of the cloud, the challenge of managing unpredictable cloud costs, and balancing customer- and market-centric decisions. You can watch or listen to the podcast or read the highlights of their discussion below.
What’s the impact of the cloud on Aerospike’s real-time business?
Lenley Hensarling: “I think in the past, very large corporations had data centers with the network bandwidth and availability of resources to handle things such as analyzing high-volume transactional data in real-time. We still have customers like that today. But part of what’s happened is a democratization of access to computing, networking, and storage resources. That’s what the cloud has brought, which has led to this huge richness of new companies providing new services all the time. We look at markets in terms of legacy companies that are doing digital transformation and modernization and new tech companies that are challengers or disruptors. Those disruptors have access to the same span of resources that huge companies do because of the cloud.
“At the same time people have learned how to use the cloud, I think they’ve learned:
- How to isolate things with Virtual Private Clouds (VPCs)
- The security models in the cloud, which are slightly different than on-prem
- Where to use the cloud and where to remain on-prem
“There’s a much more nuanced understanding of all of that.”
How have the complexities of managing budgets changed from on-prem to cloud?
Hensarling: “That’s something many of our customers grapple with. As we offer databases-as-a-service, we grapple with it as well. There’s a latency between the business benefit back to the customer and when they might be billed for things.
“People try to set parameters, but the auto-scaling model says let’s shape it very closely to what our expected usage is, and then when something happens, it may go up. When the first cloud came out, there were a lot of stories of people recognizing it as great and opening it up for use, and then they’d spend and spend and not know whose budget it was.
“So, I think we’re all learning about that. There are new tools coming online all the time to help, but it’s something that has to be modeled and taken into account. In the past, we over-provisioned and hoped we didn’t run over. There was a lot of waste built into the model. Now, there may not be as much waste, but there’s the same lack of predictability.”
In what cases would people choose on-prem over the cloud?
Hensarling: “People are becoming increasingly sophisticated about understanding the cost patterns and structures of cloud workloads. So, the cloud really excels where there’s some elasticity or you have a curve that goes up and down in terms of resource usage. You can get very dramatic savings there.
“But in a stable workload or in a workload that’s just growing linearly, sometimes you can do better managing it yourselves. We’ve seen some customers repatriate some workloads. I emphasize some, because they may have machines on-prem sitting idle and try to shift things around with virtualization and everything. But having played that game myself, you strive to get to something like 80%, which means that you’ve got 20% waste going on continuously. Why not let that be the cloud’s problem?
“People are understanding now that there’s more to the game than just moving everything to the cloud. It’s workload by workload, understanding the profile of the resource consumption, and that even leads to decisions about which cloud to deploy on, even by region sometimes.
“People are becoming increasingly sophisticated primarily because the tools to monitor, manage, and understand resource usage – and therefore cost – are getting better, both tools provided by cloud vendors and from third parties.”
What should drive product management: customers, market, or deals?
Hensarling: “I think you have to do both. A good product manager goes after a market. So, you want to provide a product that meets the needs of a fairly broad constituency of customers and, therefore, a market. But you also have to build for people who are actually going to be using it. So, really getting into a market also means getting with customers, and not just a few, but a number of them. Looking at as much data about the market and profiles of customers as possible, reading about it. When first going into financial services here, my product managers and I spent a lot of time reading and talking with subject matter experts. We worked to get a good sense of the problem sets that are common. Then, you can start to match up your requirements in your product against those problem sets.
“Sometimes companies can become not so much customer-driven, but deal-driven. ‘We want this deal, so we’re going to do these things.’ And they don’t stop and think if they’re applicable to a broader range of customers. Now, in any given situation, you might make a deal-driven decision around product if the deal is large enough or if it gives you your first foothold in a market. But what you really want to do is constantly assess what’s the broader applicability of everything you do because resources are finite. You have a choice every time you apply something. I always joke with the engineers that every line of code you write is a business decision.”
The Game = identify what’s possible and go to the adjacent possible
Hensarling: “I’d say that the product management problem is always the same: Figure out what to build to get the most money. That’s the top-level thing. That means you have to decide what to build for whom, and that means picking a market and then picking what I call canonical customers. And this all takes place within the constraints of your code base.
“I heard Stuart Kaufman speak at the first Java One conference. He’s a mathematician who does a lot of work in chaos theory, and he talked about the adjacent possible. In terms of biology, evolution can’t go anywhere – it has to go in the next space. It struck me that that had a lot to do with coding.”