Looking at Heroku's capabilities to scale Notabyte

How serverless with Heroku enabled Notabyte to scale and grow.

Developing architecture on the modern web is nothing like a decade ago and you can quickly fall into a situation where you have to rearchitect your entire application to scale and/or you hit cost barriers to growth causing price hikes in your most vulnerable stage of growth. We've been building Notabyte for years now and throughout its journey we've not hit either of those obstacles to its growth thanks to the power of the Heroku platform.

We've been able to leverage Heroku to scale our application performance and manage costs effectively to the point that it's very little in the overall tech stack overhead to operate the application itself. Our costs for Notabyte flow more from the cost of the API usage by a considerable margin (for every $20 we spend in APIs, we're spending less than $1 in the application layer). This is a massive cost savings for us and allows us to focus on the core of our application and not the infrastructure.

Not to mention having serverless applications allows us to focus on building secure applications and not have to maintain DevOps around platform security. It's not just the cost of the server, either, but also the cost of the resources required to architecture and maintain that platform. Heroku by default transitions a considerable portion of that expense off your plate by offering serverless resource hosting.

Unveiling the Power of Heroku for Notabyte

Yurika Corporation, renowned for its pioneering advancements in AI and community engagement, relies on a robust digital infrastructure to support its innovative products. Notabyte, a bot designed for social interaction and image creation, is a flagship product operated by Yurika, and at its core lies Heroku. The following case studies illustrate Heroku's tangible impact on Yurika's operational expansion and bottom-line savings, providing insights that transcend mere anecdotal evidence to showcase its real-world strength.

Navigating the Cost-Efficiency Terrain

For burgeoning enterprises like Yurika, cost constraints are an omnipresent challenge. Leveraging Heroku's serverless architecture has yielded substantial savings over conventional cloud service models. In comparison scenarios with industry giants like AWS, Notabyte's monthly running costs are estimated at roughly $2800 a month whereas with Heroku we've actualized just $50 a month to scale our high usage AI bot (it receives 10 message per second and at least one request every minute). Heroku provides a paradigm shift that affords significant financial breathing room for product development and market foothold expansion. This significant variance in operational expenses translates to a competitive edge, allowing for unprecedented scalability without inflationary pressures on the company's resources.

An example of the cost savings is that instead of a single compute layer or multiple compute layers, we operate production dynos that we can scale up and down as needed. This allows us to save on costs when the bot is not in use and scale up when we need to handle more traffic, balancing costs.

High Performance, Low Footprint

Leveraging the eco dynos for low usage components of our bot, we're able to offset costs and maintain high efficiency with the main application layer by offloading rarely used commands to a separate serverless instance. Thanks to not having to maintain the server assets, we're able to focus on coding what's best for our bot over maintaining the server. That's massive cost savings for us, just not having to operate a traditional server layer, security updates and securing an entire operating system versus securing our application.

Eco dynos, with their resource-efficient functionality, are apt for low-usage applications, ensuring a cost-effective strategy for running such services. On the other hand, regular dynos tick all boxes for enterprise-grade operations, meeting the stringent requirements of uptime and performance vital for Yurika's community and service engagements. This dynamic scalability between needs and resources underscores Heroku's adaptability, perfectly aligning with Yurika's goal of sustained excellence without compromise.

Understanding Heroku Costs

Notabyte is multiple applications replicated across what we consider a “fleet of dynos” that allows it to orchestrate its operations based on user requests. Commands like !img and !prompt run through multiple API layers for moderation and execution, which take place within a production dyno due to their high usage. Commands like !free run across multiple eco dynos, which operate generally between 5PM and 3AM ET, allowing us to utilize webhooks in seperate dyno applications without having to pay more than $5 for the total pool of resources.

While not always applicable, the eco-dynos are highly cost efficient for our needs for the small, lesser used components of our stack. The production dyno price cap of $25 allows us to properly plan our expenses each month and build that into the pricing model of Notabyte. In what a user pays 8 cents for, less than 1 cent of that goes to the actual architecture behind the bot. We have roughly 14,000 !image requests per month. That’s in total 70,000 API calls across all 14,000 requests with another 28,000 !prompt requests per month, which include 140,000 API calls. A single production instance (max $25 a month currently) handles all of this with no server errors, timeouts, concerns about RAM usage and plenty of room to continue growing. If somehow, we begin reaching a resource constraint (we’re only at about 80% right now), we can quickly expand the dyno to the next tier and continue to manage costs with their capped dyno pricing.

Furthermore, we’re able to host a secondary dyno that can act as a load balancer in times of ultra-high usage. By routing the commands between the two dynos, we’re able to again limit our costs, pay for only what we use and continue to provide an always-on and always available service.

To put in perspective, we pay roughly ~$50 monthly in Heroku expenses and $800 to $1200 in OpenAI expenses a month, in addition to another $200 in AWS API calls (for moderation, vision, voice, etc.).

Heroku enables our application costs to be almost non-existent in the total application equation.

Scalability with Heroku

All of this is maintained by a lean team without the DevOps component required to operate a VPS or Compute resource or keep Ubuntu up to date. Regarding security, Heroku’s platform is highly secure in itself, and we only have to focus on application security and the WAF (web application firewall) layer versus the traditional apt-get upgrade every day to avoid a zero-day exploit.

This allows us to continue developing with a small team a product in the market that’s got considerable usage. As our database needs expanded, we were able to increase our database volume with a quick service-add form. As we’ve begun to grow new features that don’t need to work in the main application, we’ve been able to start new dynos. Using eco dynos, we’ve been able to build webhooks that are rarely executed but vital for the growth of the bot.

The Future of Serverless

We feel that in the future, serverless is going to be more important to the enterprise and SMB space than ever before. We have almost all of our web properties now operating with a serverless vendor and are leveraging almost exclusively serverless architecture.

Our Notabyte project wouldn’t be cost effective to our users if we didn’t have the partnership of Heroku. The ability to scale, the cost savings and platform security mix together to create a beautiful orchestration of what the consumer wants where they want it at a price they’re willing to pay for.

Looking to the Horizon

The success we've had with Heroku is just the start. We're hoping to continue building on the Heroku platform and continuing to scale our application to meet the needs of our users. We're excited to see what the future holds for Notabyte and the Heroku platform. If you're curious how we can help you scale your application, reach out to us at Yurika Corporation Contact Us. We're always looking for new ways to help our users and the community at large.

Case Study

Notabyte

An AI tool for in-chat images and generative content.

Link

7 min read

Publish Date: 4/28/2024