At Momento’s public launch in November 2022, we reflected on what got us here. The “before lunch” promise in our title captured the onboarding bar we set. A developer should be able to get into Momento and issue a first set in a few minutes, rather than spend sprints provisioning a traditional cache cluster.
Our definition of speed includes more than tail latency. It includes the time developers spend getting an application to production. That is why we obsess over the power of a single API call and what serverless really means for our customers.
Our mission as founders has been to assemble an incredible team and leave a positive impact on the world. We pursue that mission by improving developer productivity and the interactivity of the applications they build. Productive developers can experiment faster and bring interactive, production-ready applications to market sooner. Consumers get responsive applications under load, which drives engagement, sustained use, and conversion.
We also obsess over the trust customers place in us as part of their critical path. That demands operational rigor. Inside the company, we value free discourse, open communication, transparency, and psychological safety. We argue passionately, and we proudly disconfirm our beliefs.
Why we focused on developer time
Watch the Momento founders video on YouTube.
We, Daniela and Khawaja, have worked together since 2014 with a shared passion for unlocking developer productivity. We have spent our collective careers building mission-critical services, making them more observable, and helping customers simplify their stacks. DynamoDB routinely serves more than 100 million transactions per second, LightStep powers observability and monitoring for some of the world’s most reliable systems, and NASA colloquially epitomizes mission-critical work.
Leading DynamoDB inspired us to deliver large-scale systems through a single API call. We wanted to hide the leaky abstractions that force developers to confront error-prone configuration, provisioning, capacity management, and outages.
Caching was missing from the serverless model
The serverless revolution preserves developer time by removing distractions such as provisioned capacity, maintenance windows, inelasticity, and exorbitant price tags. Developers can focus less on complex configurations that might cause outages or security risks and more on good code and great user experiences. Demand demonstrates the model’s appeal: developers increasingly expect serverless operations by default.
By 2022, serverless options covered compute, storage, databases, queues, and streams. Caching was the exception. That mattered because a cache is essential to many interactive applications that handle dynamic bursts of traffic.
Existing caching solutions required developers to work through numerous error-prone configurations and learn common lessons one avoidable outage at a time. A team might underscale a cache for peak traffic, fail to enable encryption, or choose the wrong instance type. At best, those mistakes hurt the user experience. At worst, they cause irreparable damage to customer trust.
The operational cost created the opportunity
Caching is one of the top five line items on the bill of almost every cloud customer we have encountered. Customers spend billions of dollars each year on caching infrastructure across cloud providers.
Building a multi-node caching cluster is time-consuming, error-prone, and operationally risky. We used spending as a proxy for the virtual machines powering caches, then used those machines as a proxy for the number of clusters. That suggested an opportunity to save customers eons of time and hundreds of millions of dollars by replacing cluster management with a single API call to a cache that works at any scale.
We did not want to build only a caching-cluster-as-a-service. We wanted to own the end-to-end caching experience, including the client libraries, a purpose-built protocol, and a serverless backend that handles the complexity of distributed systems.
One parameter replaced sprints of setup
We saw the same gap in outages across Amazon teams, AWS customers, and our own attempts to add caches to application stacks. Creating a DynamoDB table took one API call, while adding a cache took sprints. Multiplied across every cache, that setup time distracted developers from experimenting with and improving their core business. We saw nothing else on the market come close to solving the whole problem; managed services offered incremental improvements to legacy caching solutions.
At launch, Momento was the first truly serverless cache. It required one parameter: the cache name. With one API call, customers could provision a secure cache that scaled to millions of requests per second, with high-availability practices built in.
The service also used a radical, one-dimensional, pay-per-use pricing model that did not require mathematical gymnastics. It avoided the familiar perils of provisioned capacity: pay too much when you overprovision, experience an outage when you underprovision, or suffer both. The goal was a robust cache that arrived fast out of the box, with low tail latency, high availability, and security best practices that required no configuration.
Production customers put the idea to work
Momento was ready for production at launch. Customers including Wyze Labs, Paramount, NTT DOCOMO, and many others used our highly available, instantly elastic, high-performance cache for significant workloads. It was fulfilling to see those customers alleviate their caching problems and share our passion for improving the end-user experience.
Watch the Wyze customer spotlight on Vimeo.
Our engineers brought experience building large-scale, mission-critical, performance-sensitive systems at Google, Amazon, Valve, Intuit, and GitHub. We held on to the operational rigor that helped us support mission-critical workflows at larger firms. Momento customers valued our obsession, responsiveness, and relentless focus on operational excellence.
Keep the onboarding bar high
At our November 2022 launch, watching a developer try Momento Cache Serverless for the first time was inspiring. We set the onboarding bar for that Serverless experience at less time than provisioning a traditional cache cluster, and we routinely measured it. Our invitation then was simple: try it. Let us know if getting into Momento Cache Serverless and issuing your first set took longer than a few minutes.
At that launch, Momento Cache Serverless had arrived. We had an incredible team, a service customers were already falling in love with, and a vision to make the world more productive. Our conviction deepened each time we pushed a new customer to production. We described the launch-era Serverless multi-tenant architecture as having fundamentally different economics, better availability, and instant elasticity unmatched by legacy solutions.
Our people-first culture and our team’s customer focus reinforce each other. Best of all, we are just getting started.
