Interview with serverless apps monitoring startup Epsagon

Nitzan Shapira is the Co-Founder of Epsagon, an Israel-based Startup focusing on an AI powered monitoring solution for serverless applications.
We — superluminar — met him at JeffConf Hamburg. superluminar is based in Hamburg, offers Cloud Consulting, loves serverless and hosted JeffConf Hamburg.

Superluminar: Hi Nitzan, please introduce yourself!

Nitzan: I’m from Tel Aviv, Israel. I hold a M.Sc. in Computer Science. During the last 12 years I have been working as a software engineer, researcher, and R&D team leader — in Intel, Israel Defense Forces and a startup.

Superluminar: When and why did you get in touch with Serverless for the first time?

Nitzan: My partner Ran was already experienced with developing skills for the Amazon Alexa platform, so he was familiar with serverless. During the last year we have been exploring the serverless domain. Both Ran and I have many years of experience with infrastructure technologies, cyber-security, and R&D, so we felt comfortable researching challenges in cloud infrastructure and serverless.

Superluminar: What was the key moment to found Epsagon?

Nitzan: I can say that it wasn’t just a single moment but a process of several months, in which we have interviewed many customers (serverless users) in order to understand their pain points, and the ways that they address them. After many such interviews we came to the conclusion that serverless users today cannot understand what is really going on in their production environment, due to the complexity of serverless architectures. This leads to major challenges such as monitoring, troubleshooting, and more. That is when we decided to address this challenge.

Superluminar: what’s the problem in monitoring serverless applications you want to tackle with Epsagon?

Nitzan: the core problem comes from the nature of serverless architectures. Since there are suddenly so many elements: compute units (e.g. functions), cloud resources and external APIs that the architecture is utilizing, these architectures are highly distributed and event-driven. Therefore, while it is possible to analyze a specific function, it is very difficult to obtain deep insights about the end-to-end performance of the system, as well as to troubleshoot errors. The key challenge is that all of these asynchronous events must be connected somehow — and it must be done automatically.

Superluminar: Btw: We interviewed JeffConf Hamburg attendees and I think this result looks promising for Epsagon:

Nitzan: I am not surprised. This matches our conclusion after many customer interviews. And we keep seeing this feedback nowadays when speaking to new customers.

Superluminar: You raised $4.1 Million as a stealth startup, that’s impressive. Just curious: how did VCs react as you told them about your Serverless monitoring plans or did they just focus on the AI because Serverless is still at the beginning in their opinion?

Nitzan: it is true that serverless is a new field and in many ways still in its early days. However, it is very clear that major players (e.g. cloud vendors such as AWS) are pushing it very hard, as well as innovative enterprises which are already utilizing it in mission-critical production systems in a very high scale. So even though it is new — it is already here. Many VCs today believe (as do I) that serverless is going to be the default way to design and run applications in the cloud in the future. The AI was actually not the main thing — the main issue is that 1) serverless is going to be huge, and 2) people are having major difficulties with monitoring in serverless TODAY. Therefore — this is going to be a very big issue in the near future.

Superluminar: Back to your product: what does your private beta offers today and what comes next?

Nitzan: our private beta, which is running is production environments today, enables users to understand what is happening in their production environment. It automatically identifies all the elements in the system including functions, databases, queues, storage elements, external APIs in use, and more. Our algorithms provides visualization of the entire serverless architecture, and traces asynchronous transactions from end-to-end. This enables very fast troubleshooting for errors and issues, and the understanding of performance of different flows in the system, which allows to identify bottlenecks — such as an inefficient Lambda function, or a slow API that is in use. We are currently working on adding support for more and more APIs, as well to further develop our AI to identify issues in the system that today, due to the lack of visibility, people are not even aware of. We support the main programming languages that people use in serverless today.

Superluminar: If someone is new to serverless computing, what would recommend to him or her?

Nitzan: start with reading some documentation, but more importantly — just start deploying functions and utilizing cloud resources and APIs — you will be surprised how easy it is!

Superluminar: Complete the sentence: Serverless in 2020 will be…

Nitzan: the default way to design, deploy and operate cloud applications.

Superluminar: Thanks Nitzan! Hope to see you soon again!

Nitzan: Thanks! It was a pleasure. I hope to visit Hamburg again soon!
Our private beta is live, and I will happily answer any questions to anyone who’s interested in the product.