ZigZaGame: How we launched Tokyo Debunker with low-latency messaging without WebSockets!

ZigZaGame is a global mobile gaming studio that released its latest title Tokyo Debunker in April 2024, on the heels of their breakout title Evertale (10M downloads in 122 countries). Momento Cache and Topics, our low-latency event bus, helped to accelerate and scale ZigZaGame’s backend so they could focus on what they do best — releasing new features and titles that delight their fans.

We talked to Tom Delorenzi, CTO of ZigZaGame, about his experience launching Tokyo Debunker with Cache and Topics.

Can you tell us a little about the technical stack behind TD?

‍We built our entire platform on AWS — and we are a serverless first company. We’re using the Momento platform — both Cache and Topics. Together, these two services underpin the platform that powers TD’s interactive experiences.

How did you start thinking about implementing Momento Topics for TD?

We’d implemented Momento Cache for our previous title (Evertale), and it worked so well we also decided to apply it to Tokyo Debunker. This time around, it was a new choice to replace WebSockets with Momento Topics. Tokyo Debunker has several gameplay features that rely on low-latency interactivity, and we wanted to drive them in a better way without being distracted by WebSockets management. 

Why did you want to replace Websockets?

WebSockets is how we powered chat in Evertale, but it required a fair amount of subscription management code (Lambdas) that we had to write and maintain. This is expensive operationally (and on the infrastructure). 

Our original architecture with WebSockets had cursory API Gateway and Lambas for bookkeeping connections.

Evertale (Original architecture)

Momento Topics meaningfully simplified our architecture. Momento owns the client side SDKs, the subscription management, the authentication, the API layer, and the message bus. This batteries-included approach allows us to simply publish and subscribe to messages without any of the underlying nuances.

Tokyo Debunker (New architecture)

There are a few big differences between our pub/sub for Evertale and for implementing Topics on Tokyo Debunker. With Evertale it was a “pull” system, whereas now with Tokyo Debunker, it’s a “push” system. Topics “pushes” messages down to clients. The flexibility on topic authentication on a per key prefix and the client being able to get their own data opened up a lot of possibilities for reducing server workload by offloading data fetches to the client.

Overall, migrating to Momento Topics from our previous architecture was easy. It took us only a few days to write the main code. We did some learning with the permissions, tokens, limit increases, and timeouts over a few months of testing. In the end, we tuned alongside the Momento team to prepare for a seamless production launch of Tokyo Debunker.

With the success of the Raid feature [that we launched in April 2024], we’ve now applied Topics to the War feature (“Game of Claws”), which is a mass combat, high pub/sub scale feature with real-time gameplay. 

• Players form Guilds, and together they participate in a two-week long war twice a month made up of real-time battles.
• Guilds battle for occupation of castles (filled with the game’s distinctive collectible cats), and they can either defend or attack castles to receive rewards. At the end of the 2 weeks, the top Guild wins by accruing the most points.
• This is a lot more high stakes for players than the Raid feature, and they stand to gain a lot more rewards.

This is also a much more intensive use of Topics — the battles all occur in a concentrated, single 45-minute period every day. Unlike Raid, these war battles are real time, expecting an action every 2 seconds and are very unforgiving on latency. We consider it a mission critical, true real-time test of Topics. We went live in July [2024], and it has gone very well!

In the interim, there are also a few distinctive differences that we’ve already noticed between Evertale and Tokyo Debunker based off of some common features using Topics.  

• Both Evertale and Tokyo Debunker have a Replays feature, in which players can choose to replay a battle). On Tokyo Debunker with Topics, players can draw replays directly from Momento with lower latency than on Evertale. 
• The architectural changes will also benefit the Guild War feature for Tokyo Debunker, compared to Evertale. 

How has Momento tech impacted how you and your game dev team think about future game titles?  

At the end of the day, my game devs are focused on building the best game they can. As an engineer, I find solutions to make the game devs’ vision possible, and I’m always searching for better ways to do them. Momento tech has given us new ways to do things better, and with these new capabilities (and time and money saved), it can help us create new ideas for our games in the future. Next year, we’ll release another title — and we’re planning to continue building on the Momento platform. 

I consider success with Momento measured primarily on cost and reliability factors. For cost, it’s proportional to usage instead of capacity, and for reliability — if Momento can remain as reliable as it’s been so far, we will continue to be happy customers. 

How did you first start working with Momento? 

Last year (2023), our SA at AWS, Matt Nimmo, suggested looking into Momento Cache because we had experienced failures with our connection spikes. At the time, we were on board with becoming a serverless first company, and we had gone all in Lambda because we enjoy the pay-per-use pricing model and elasticity to handle seasonal spikes in our game. However, moving to Lambda was tough: ElastiCache Redis struggled to keep up with the connection spikes from Lambdas — and we risked disrupting player experience during those spikes. 

We deployed Momento Cache on Evertale’s in-game chat system, and Momento’s routing layer absorbed our connection spikes — and it offers the consumption based pricing model that we loved in Lamba. Momento Cache integrated seamlessly into our stack, absorbed connection spikes, and it just worked. It took just 3 weeks to deploy. 

Do you have any advice for other gaming studios on how to start working with Momento? 

Yes, I’d recommend starting with a more confined, low-stakes workload for the POC. For instance, we first implemented Momento Cache on Evertale (our older title) for the in-game chat system so we could mitigate the risk of impacting broader gameplay. 

In the end, Momento Cache and Topics has been so reliable and easy to use that we’ve decided to use it for future titles, including Tokyo Debunker and our new game title planned for 2025.

• ‍Reach out to us for a free consultation — Our solutions experts are on deck to share more about how Momento Topics can help optimize the real-time interactive features in your game or app. We’d also be glad to talk through ideas for POC’s. 
• Sign up to try Momento Cache or Topics for free — create an account 
• Play Tokyo Debunker — Google Play or Apple