Podcast: Episode 5
Download our guide to learn how streaming businesses can optimize their architecture to save costs.Download now
From its simplicity to its scalability, serverless computing offers businesses an affordable and efficient way to make developers’ lives easier without compromising results. But how does serverless work? And is it everything its boosters say it is?
In this episode, host Rachael Lewis-Krisky talks to Anshu Agarwal (she/her), DigitalOcean VP and GM of Serverless, about all things serverless to learn how it works, what it can do, and how your business can make the most of it.
Anshu is the general manager of Serverless Business Unit at DigitalOcean. Prior to this, she was the co-founder and CEO of Nimbella, a serverless cloud provider. She has over 20 years of experience in the technology industry, building and evangelizing products that have solved real-world problems for organizations large and small across the globe.
Watch Anshu speak more about serverless with TFiR
[00:00:00] Rachael Lewis-Krisky: Welcome to ‘Making Work Work,’ the DigitalOcean Podcast, where we explore everything about work - the tech that makes it possible and the people that make it happen. And of course, that’s never as simple as it sounds. I’m your host, Rachael Lewis-Krisky.
Serverless is kind of a big deal. It’s changed the way a lot of businesses are approaching their infrastructure. It’s simple and scalable, but as is usually the case for new technology, it can be a little confusing.
How does it work? Is it everything it’s cracked up to be? In this episode, we’re talking to Anshu Agarwal the Vice President and General Manager of Serverless here at Digital Ocean. We’ll be discussing the ins and outs of serverless and if it’s right for your business or not.
Let’s dive in.
Anshu Thank you for chatting with me on making Work Work.
[00:01:19] Anshu Agarwal: Thank you, Rae. It’s a pleasure to be here.
[00:01:21] Rachael Lewis-Krisky: I am really excited to have you on our show to talk about serverless computing in particular because of your background from facial recognition with artificial intelligence to contact delivery networks, experiences with several startup acquisitions to serverless technology, and then eventually joining DigitalOcean.
You’ve had such an amazing career in technology spaces so far.
[00:01:42] Anshu Agarwal: Yes. I’ve been actually very fortunate being part of five startup journeys, uh, and last one being my own, and then getting acquired by DO. It’s been, it’s been a wonderful experience so far and I’ve enjoyed every part of my career.
[00:01:54] Rachael Lewis-Krisky: Absolutely. So for this installment of what’s next in tech, we’re talking serverless; exploring what it is, how it fits in to work that businesses and developers are doing. Why is this, what’s next in tech? Is it what’s next in tech?
[00:02:09] Anshu Agarwal: Well, yes it is. Cloud is inevitable and we have proven it now. Companies who thought they would never move to cloud often end up fully operating in the cloud. And any new company that is born today is born in the cloud.
So today there are 27 million developers around the world, and this number is not stopping here. It’s going to grow exponentially. Quite a large number of these developers have less than two years of experience, and they don’t have time to learn the cloud. So the cloud has to become simple. Businesses today have to deliver value to their customers at the pace of innovation and not spend time learning infrastructure because it doesn’t add any business value.
So what’s next in tech is the methodology that simplifies and democratizes cloud for everyone. This so-called supercomputing power of the cloud is available to the developers of all skill levels and not to just cloud experts. This is what serverless promises; the promise to relieve those developers from all infrastructure burdens. So the developer only has to work on adding business value and getting to market faster.
[00:03:15] Rachael Lewis-Krisky: And you said democratizing. Now with technology, that’s a very interesting terminology that usually is applying to civil and social spaces. But how is that unique to serverless and the tech spaces?
[00:03:26] Anshu Agarwal: So what democratization means in this context is, when cloud started, it had one or two services, simple to understand, easier to move forward with than doing your own servers and everything.
Then we added so many services to the cloud and made cloud very difficult place for most, people, most developers. Adding those services gave them more, things to do in cloud, but added a lot of complexity. And then we added cloud certifications. So that took away the benefit of, you know, cloud is for everyone. It became cloud for select few cloud experts and everybody spend time learning the cloud.
What democratizing cloud means is that it is a cloud for everyone. Anybody, even if they’re a high school student, or even if they’re a middle school student, they should be able to use the cloud the way they would use anything that is simple, like more like we used to think of plug and play appliance. That’s what democratization of cloud means.
And what serverless does is you don’t have to learn infrastructure. You don’t need to know how to scale your application, how to make your application reliable. All you need to do is learn how to code your business logic, and the application takes care of everything else. That is what serverless promises. Where you are spending all your time on logic of your application or your business, and no time on infrastructure. No time learning the infrastructure, no time scaling the infrastructure, no time maintaining the infrastructure. That’s what we mean by democratizing the cloud and giving the supercomputing of cloud to everyone.
[00:04:58] Rachael Lewis-Krisky: And you bring up that it makes it so anyone can do it, even middle schoolers. I know you have a very unique experience with that.
[00:05:04] Anshu Agarwal: Yes, absolutely. I have a kid, uh, who is now in college, but he started using, cloud services when he was in sixth grade. And he developed applications for entire sixth grade, so sixth graders could understand how to use cloud. So he, he did that. Living in Silicon Valley, you can’t escape tech and you can’t escape from the kids who actually want to experiment with tech at a very young age, and it’s, it’s, wonderful.
[00:05:30] Rachael Lewis-Krisky: I love hearing that there was democratization happening in Silicon Valley.
So before diving into nuances of adapting to new technologies for businesses that are maybe exploring serverless for the first time, or even, developers that are exploring serverless, I do wanna better understand just what it is. So from a technical exploration standpoint, where did it come from? How does it scale with traffic? Can you sort of, uh, give me the, 101 pitch?
[00:05:54] Anshu Agarwal: Yes. So I’ll have to, uh, very brief cloud computing history lesson. Okay. Throughout the cloud computings evolution, we have had multiple forms of abstraction levels. The data center, whether it was an on-premise or provided by a vendor as a service, was the first step to the cloud as we know it today, it abstracted the physical hosting environment and we started scaling those environments with hardware units.
As virtualization evolved, we started hosting virtual machines on cloud environments, and we abstracted the hardware and used operating systems as the unit of scale. Soon after that, we started building hosting environments to the cloud and abstracted the operating system, which became our new scaling unit, was application. And now we move to serverless architecture that allows us to build small pieces of code that do something useful and at the same time run quickly without consuming large amount of server resources.
All of these cloud models that I talk about in the evolution leave us with different responsibilities. Having the data center on premise meant we had the complete responsibility over everything. As we move to the cloud, each step in the evolution left us with fewer and fewer responsibilities. And with greater level of abstraction comes with greater level of automation.
So, long story short, you know, how does it work? How does it scale? Serverless allows applications to scale on demand through concurrency, while consuming very, very small amount of resources. Compared to server proof computing. It is the best way to scale because you’re never over provisioning.
[00:07:26] Rachael Lewis-Krisky: It’s very easy to have preconceived notions around what different aspects of tech is–you hear serverless, the immediate assumption is there’s no servers, but that’s ridiculous. So, I think about if people are listening and they’re wondering if it’s feasible, if it’s scalable, is it easy to learn a lot of these questions, um, around whether or not it’s a good choice for them? Can you sort of talk about what the reality is of serverless that might affect people’s decision to dive in with it?
[00:07:51] Anshu Agarwal: Yes, absolutely. Uh, one of the one that you said that serverless means no servers. No actually there are a lot of servers, but those servers are managed and maintained by the cloud provider. So you have nothing to do with the service. You don’t have to do any upgrades, no patches, nothing at all. So serverless is serverless for the user; not for the vendor. Okay.
The biggest misconception of serverless is that serverless equals FaaS. And, you’ve heard that acronym in many articles. FaaS stands for function as a service. FaaS is only the compute part of serverless and many other services out there, which provide additional functionality like serverless databases, API gateways, and more, that actually constitute the serverless methodology. So a serverless methodology is what enables you to not worry about managing infrastructure, not worry about scaling because it’s available out of the box to you and you pay only for what you use. So that is like the biggest misconception that serverless is FaaS. Serverless includes FaaS, but there are so many aspects of serverless.
The second misconception I would say, which is pretty big, is the time I save by not managing infrastructure will be replaced by supporting complexities. So some folks think–who have not used serverless– that serverless is complex. Because the architecture becomes complex. With a serverless approach, the time working on managing updates, patching, security, et cetera, et cetera, will absolutely decreased. That is given. The whole point of adopting serverless approach is so more time can be spent impacting users of that infrastructure by helping them to build more robust and reliable solutions.
So, when we break down an application into smaller pieces, we are creating much more flexible infrastructure, and you can develop independently of what used to be one core application. So now the application you work on starts to have more straightforward integration points and it scales as needed. This simplicity of integration, of serverless pieces is what makes your platform thrive. So the misconception that it is complex is completely wrong because you are actually simplifying it for the future.
[00:10:10] Rachael Lewis-Krisky: Your knowledge based on this particular subject area is robust, obviously because of your experience with Nimbella which you founded four years ago, correct?
[00:10:18] Anshu Agarwal: Yes, that is correct.
[00:10:19] Rachael Lewis-Krisky: I’m always interested in how people sort of knew they were onto a good idea. So I’m curious if you could talk to us about, how you knew you were on the right track with serverless.
[00:10:28] Anshu Agarwal: That is a great question. So when we founded Nimbella, serverless was popular among certain types of developers. But they were all doing simple stateless applications. The applications that didn’t rely on any kind of persistent storage. We saw that all serverless frameworks that were running in major providers were proprietary and you couldn’t build a stateful application, easily without complicating your architecture.
So whenever you start a company, you have a conviction in the problem, and once that conviction is there, you find a solution that solves the problem. And this is how we founded Nimbella. Developers would build using an open source framework so they are not locked in with a cloud vendor and they could run any stateful application without complicating their architecture. That was our differentiation. That was our thesis.
[00:11:16] Rachael Lewis-Krisky: What was a unique challenge that you faced in your time founding and launching Nimbella?
[00:11:22] Anshu Agarwal: Starting a new company is very challenging. Okay. And there are many, many challenges. There were some unique challenges I would say that I particularly faced. uh, so when we first started, hiring was a big problem. You are based in Silicon Valley hiring alongside Facebook, Google, is very hard because a startup doesn’t pay as much, first of all. The perks that are available in these companies are just incredible. No match, For any startup.
And then, you know who predicted pandemic? Nobody. Okay. So when we started the Nimbella we actually haven’t even released our product and pandemic hits us. For a new company, a new tech is always hard to establish during uncertain times, unless you are dealing with tech that handles the problem of the day. So if you were in web conferencing, it was a no-brainer tech at that time when Covid started.
So that was a big challenge. Covid hit. We are a new company. We are venture funded. So how, how do we make sure that we can weather that storm.
And, uh, lastly, but not the least, although we have made a lot of progress, having women founders, even in Silicon Valley, it is a challenge. Raising money as a women founder and CEO, for a tech company is hard. And you would be surprised that a bar for a woman’s CEO is much, much higher than the men counterparts. So that’s why I said There were unique challenges with the time and as well as who I am and in the place I am.
[00:12:48] Rachael Lewis-Krisky: But you overcame those with flying colors. You were acquired by DigitalOcean, which means that the services and the product that Nimbella was creating was available to so many more people as well. That must have been very exciting.
[00:13:00] Anshu Agarwal: Oh, it was very exciting actually. We were featured in an analyst report as the only private vendor with a serverless offering that can compete with a hyperscaler offering. So it was it was amazing. Yes. Journey was amazing, I have to say. Mm-hmm.
[00:13:14] Rachael Lewis-Krisky: How is Serverless different on DO versus other providers? Or is the technology relatively homogenous across industries, but it’s just the unique offerings and approach to serverless that changes?
[00:13:25] Anshu Agarwal: What you can do with serverless is relatively homogenous across the industry, but how you do it is different. DigitalOcean serverless is based on open source with a large community behind it, so your code doesn’t run on any proprietary platform.
Actually my co-founder at Nimbella was the creator of Apache OpenWhisk, and he’s part of DigitalOcean functions team and that’s what we have used to create DigitalOcean functions product.
Another important difference, I would say, is rooted in DO’s core tenant: simplicity. And we have adopted simplicity in both product and pricing. We have made this journey like a guided tour where you can go from discovery to development and then straight to production. The Functions cloud console is for getting started and introducing functions to the developers who are very new to serverless, because serverless is a computing paradigm and people have to learn and adopt that paradigm.
So we kind of introduce it in a form of a guided tour. Once you gain familiarity with the paradigm, you can test your functions before pushing it to production. You can use a command line interface to develop locally with your favorite IDs and then ship quickly to the cloud. Then you can grow from development of functions to fully managed apps using app platform by committing your functions to GitHub to be deployed to an app in production. You’re not learning a cloud, you’re not learning the ways of the cloud. And that’s what is, it is extremely simple.
[00:14:44] Rachael Lewis-Krisky: And we do love simplicity at DigitalOcean don’t we?
[00:14:47] Anshu Agarwal: Yes, we do.
[00:14:48] Rachael Lewis-Krisky: Related, but switching gears a little bit, in previous interviews you’ve talked about how businesses can get stuck, sort of drinking the Kool-Aid, right? Believing that their product and their company is almost perfect. Why is adaptation so hard for businesses?
[00:15:02] Anshu Agarwal: Well, the thing is, if it’s your own baby, so nothing can be wrong. with it. That’s, that’s the feeling. This is how we feel about our products. However, it may not be the case most of the time. It is hard because when we look at a problem, we look at our solution, we look at from our own lens. And we cannot see beyond it.
So talking to customers and listening I found is the most important thing to get out of this mindset. I think listening, being aware of the market and understanding why a customer churn. is how you adapt and grow. It is sometimes more important to know why you lost a customer than why you gained a customer, because you will learn more on what you’re not seeing. Why you gained a customer is what you are already seeing. But what you’re not seeing is why you lost a customer.
And I have been a product person for a better part of my career, and one thing I’ve learned that if you don’t cannibalize your own product, someone else would come and cannibalize it for you. You have to keep your eyes and open to that, and obviously this is one extreme, but I’m always thinking that what can replace my product and how can I be a better replacement of my own product so I can future protect my customer’s interests?
[00:16:14] Rachael Lewis-Krisky: That’s a very, constructive thought exercise. You know, it challenges you to take yourself out of your daily work and your pride even to actually think about how you can make a bigger impact and do your work better.
[00:16:26] Anshu Agarwal: And as a product leader, you have to always think about it because as an engineer you’re kind of developing, you know, this is how I’m going to do it, and this is how, this is the best way. But as a product leader, you always have to think, about things that actually are challenging. And sometimes you feel that, you know, you are going in a wrong direction. But that feeling is important.
[00:16:44] Rachael Lewis-Krisky: So if a company has to consider whether or not they need to adapt to serverless, how can they actually consider if it’s the right fit for them? You know, new is not always better. So is this really good for every company?
[00:16:57] Anshu Agarwal: So serverless may not be suited for all customers’ workloads. If your workload is constant and very long running, you may not realize all the benefits of serverless. Uh, so it may be better to stay with serverful implementation.
Cloud is an evolution. It doesn’t mean all applications have to run it. There are applications for VMs, there are applications for even direct servers, okay, completely bare metal. But you should be able to decide what is suited for which phase of the cloud evolution. So new is not always better for everything, but new is always trying to make things better. Otherwise there’s no reason to do it.
[00:17:35] Rachael Lewis-Krisky: So based on. that, if it was clear to someone, let’s say, a developer on a team at a small business that this actually would be a good fit, how would you suggest that they discuss this as an option in their department? Is it as simple as, Hey, let’s look into this? Or do you recommend a more thorough pitch? Is leadership, you know, apprehensive about these kinds of changes? What might that look like?
[00:17:56] Anshu Agarwal: It is very easy just try it. Okay? It doesn’t a picture or anything, just try it out. For example, go to DigitalOcean Console, sign up for functions and try out a sample repo. Seeing how easy it is to build, deploy, and iterate is what is most important to look at. And moreover, it comes with free allowance. You’re not paying for anything for trying it out. That’s how you get started.
For real workloads, I would suggest starting like a new app with serverless architecture. Not changing an existing app, but starting a new app, and then migrating your existing apps. Most developers migrate to serverless in stages. And the ability to combine long-running server based apps with on-demand functions is very important. It is perfect for extending existing applications with new serverless functionality and doing so with speed and agility of simply writing small functions.
We have a term that we use here, uh, at DigitalOcean for this; we call this general migration, and we offer it through one unified experience of building serverfull and serverless applications.
So it doesn’t require a pitch to leadership team or anything. And having sold to many companies the solution, developers in today’s world decide what to use because they have to divert, do the work, they have to get productive on developing something. So they decide what’s best. And how they decide is by trying it out.
[00:19:23] Rachael Lewis-Krisky: So for this chapter, we’ve explored serverless, but I’m also curious about what’s next in tech that you’re actually excited about, aside from serverless. What else is there?
[00:19:32] Anshu Agarwal: At this point in time, even today, you still need to be a developer to use technology. What has changed is the skills you need to use then–we’ve simplified it. We’ve simplified to the point that you don’t need to be an expert in using some stuff, but you still need to be a developer.
There is still a lot of work that needs to be done in developer experience, so someone who doesn’t know any computer language that can do the same things that a developer does. I think AI, ML, NoCode and many other techs are building on current stage of the cloud, and probably the future is that anyone without any knowledge of computer languages would be able to build applications because complexity will be absorbed in the cloud, and the dev experience would be extremely simplified.
Don’t get me wrong, though, there will always be need for expert developers because someone has to create these platforms, but everyone will be a developer without knowing the inner working of cloud or the computer.
When we started this session, we said there are 27 million developers today. Who knows, there could be 270 million developers in no time if we can offer that kind of simplicity on the cloud.
[00:20:35] Rachael Lewis-Krisky: And that’s a true exponential technology, isn’t it?
[00:20:37] Anshu Agarwal: Of course that is, yes. And cloud evolves and cloud has been evolving uh, since we have known the cloud. So I’m pretty optimistic
[00:20:44] Rachael Lewis-Krisky: Though ‘Making Work, Work’ as a show is relatively new. I have a question that I like to ask all of our guests as we close out the interview. We do so much in our lives besides actually working –besides working on serverless in this case– which informs what we do on a regular basis. So I’m curious to know if there’s something you’ve read or watched or listened to, even that might be seemingly irrelevant from your day job that actually has kind of informed your perspective at work or helped you solve a problem at work. Can you talk about that?
[00:21:13] Anshu Agarwal: That is a difficult question, Rae. I wish I I, had time to read more because I think, reading is a skill I lost a couple of years ago. Everything is in bite size pieces I absorb through online articles and stuff like that.
But, uh, a friend of mine just gifted me a book the title of the, book is ‘Why We Sleep: Unlocking the Power of Sleep and Dreams’. It’s by a professor at Berkeley, and it is about sleep, we all know sleep is important, but you know, what do we sacrifice most when we are in a bind? We have to finish something. We have to finish a project. When we were students, we had homework to you. whatever. throughout our life, we have been compromising on sleep.
And that is one thing a lesson that I’ve learned is that sleep is probably the most important thing that we shouldn’t com we shouldn’t compromise on no matter what it is. Because it impacts on how we work, what we work on, how productive we are, what ideas we generate, and everything. And health is a given, but even our work and how effective we can be at work, what kind of ideas we can bring. They’re all developed in sleep.
I would highly recommend reading it because it gave me a new perspective on what we should be doing to be productive And this is something I would just say I have learned having been part of so many startup journeys, all I have compromised is sleep. It not only helps my health and body, it helps my work. That aspect I didn’t realize because at some . Level I knew, you know, I’m not very productive, but understanding the science behind it was so important for me.
[00:22:56] Rachael Lewis-Krisky: It literally made you change your approach and your habits to your body and your health. I have no doubt that that will resonate with a lot of people, including myself,
[00:23:05] Anshu Agarwal: Excellent.
[00:23:06] Rachael Lewis-Krisky: Well, as lovely as our conversation is, we can’t talk forever. I really appreciate you joining me on the show. It was fascinating to learn that despite serverless being an advanced technology, it’s actually fairly simple for developers and businesses to work with. This has been ‘Making Work Work.’ Whether you’re a listener or a customer, the work we do wouldn’t be possible without you.
Special thanks to all the people that make my work work, my DigitalOcean colleagues and friends.
Our music is composed by Mirco Altenbach.
And on this episode, this end credit song is by DigitalOcean’s Matt Anderson, a Security Operations Center Analyst.
Our employees are diverse in experiences and talents, so we’re excited to bring you a unique original song from one of our sharks nearly every week. Hang out and listen to the rest.
Check out do.co/makingworkwork for more info.
Until next time, I’m Rachael Lewis-Krisky. Keep swimming friends.
Sign up now and you'll be up and running on DigitalOcean in just minutes.