In part 3, Hillel stated, “Let’s talk about Google Cloud because, obviously, Amazon has been largely dominating the serverless application landscape. I think people are using Google Cloud functions, and they’re using Azure functions. At least what we see — is when you look at production applications that have really been designed serverless ground up, the vast majority are still running on AWS. But Google did have its yearly Cloud Next event in San Francisco two weeks ago, where I actually had the opportunity to present our alpha version of what we’re doing there on stage, which was nice. If anybody out there wants to preregister for our beta program, that’s great.
“But one of the interesting announcements at Google Cloud Next was this Google Cloud Run platform. I really wanted to pick your brain, Ran, and how you see this. The way I understand it, Cloud Run is essentially kind of a serverless container platform, so it’s giving you a lot more control over the runtime. You get to pretty much pick whatever Docker container you want to run there. You still get something that kind of feels like serverless, but not necessarily single event per container. It could be multiple events per container. Maybe you do have to think a little bit about scaling and things like that. Where do you see these things on the roadmap? Obviously, there’s Fargate and AWS and now there’s Cloud Run. Where do you see these pieces on the serverless application landscape? Are these important pieces? Are they blips? How do you guys approach it?”
Where to Place Things on the Serverless Spectrum
Ran stated, “It’s definitely important and I think just when AWS released Fargate and people didn’t know how to react to it. Is it serverless? Is it managed, not managed? They didn’t know how to put it on the spectrum of serverless and they placed in somewhere between ECS and a Lambda function. I think Google Cloud Run is somewhere between Fargate and a Lambda function. You’ve still got some awareness of this notion of operating system and a docker file and everything, but it reduces the management, overhead, even more than Fargate. I mean with Fargate you’d still have to manage how much memory and instances and so on, but it’s even less than that.
“I think it’s crucial because, as I said before, serverless will adapt to our use cases and this is one scenario where, for example, I want to have maybe a single container always warmed up because I want to serve this HTTP request blazingly fast or as fast as I can, but I still want zero or close to zero management of infrastructure because the serverless notion is more about reducing infrastructure management, rather than scale to zero or don’t pay if you don’t use it. I really don’t care. I’m okay with having one instance that will cost a really small amount of money, but making sure that my request will handle faster if it takes a bit more money.
“This is one scenario where we’re starting to see there’s a really gray boundary between services one to another and it’s more about the spectrum where how far you want it to be managed for you and how well you’re willing to compromise maybe security or maybe performance or maybe cost. It’s really a tradeoff where you can place yourself in the serverless spectrum.”
Hillel replied, “Yeah, I think I agree with most of that. If I ask you very specifically in terms of Epsagon, let’s say I’m already an Epsagon customer, but I might be looking at Google Cloud as a platform I want to have stuff on, and I might be looking at Fargate or Cloud Run that I have stuff on. Are these areas where I can expect Epsagon to tackle these over the next year or so?”
Ran said, “Definitely. We started, obviously, with AWS Lambda because it makes more sense, as you said. As everyone knows, AWS almost dominates the serverless market, specifically if we’re talking about function as a service. But as of now, we’re already growing to more microservices orientation or microservices that run on modern environments, like any cloud vendor. As of now, we’ve got some customers running workloads that have nothing to do with Lambda functions and they’re running on ECS or EC2 or Fargate. We don’t yet have any customers that are running on or any specific VEN2 support. Let’s say Google Cloud at the moment or Azure, but in the next year, it will definitely be. In general, I think that the way we’re going through is more about the cloud agnostic and more framework agnostic. I don’t care if it’s a Fargate or Google Run or a Lambda function. The Epsagon tracer will collect the right trace and we’ll show you this distributed tracing that spans over multiple resources.”
Read and watch part 4, “DO ALL THE CLOUDS!”