Blacksmith builds faster CI analytics with Tinybird

Blacksmith improves developer velocity by replacing outdated GitHub Actions compute with cutting-edge gaming servers, delivering 2x faster workflow runs at half the cost by changing just a single line of code.

Developers have been running CI in the dark

Throughout their software careers, the founders at Blacksmith saw a concerning trend: Developers using GitHub Actions didn't have visibility into the performance and cost structure of their CI pipelines.

They were determined to build a solution that would not only improve the performance of CI runs, but also give developers the insights they need to understand things like how workflow runtimes are trending over time, how different commits impact workflow performance, and which changes were influencing flaky test rates, among other things.

First Postgres, but then...

As with many early-stage startups, Blacksmith built their core data infrastructure on top of Postgres. They began exploring ways to build CI analytics on top of the Postgres database, but that introduced two problems:

1. Poor analytics performance. Even at their current scale, Postgres demonstrated suboptimal performance for analytics. The Blacksmith team knew the importance of speed and its impact on user experience, so growing query latencies were concerning. They knew vanilla Postgres wouldn't keep up with their forecasted growth.
2. Running analytics on their production database was risky. Running analytics on Postgres could potentially impact their application performance, so they knew they'd need to move their analytics compute to a separate instance to avoid slowing their users down.

Aayush and his team evaluated multiple options. They could invest in Postgres extensions to improve performance. They could instantiate read replicas to isolate data analytics away from their production database. Or they could build their own analytics service by self-hosting the database, building an ETL service to extract data from Postgres into this new platform, and scaling it as needed.

None of these options made sense.

Blacksmith chose Tinybird for speed

After deciding to use a managed analytics service rather than build their own, Blacksmith evaluated Tinybird and ClickHouse Cloud, among others.

The goal was simple: avoid building any infrastructure, including the database and surrounding services like write buffering and backend APIs.

Aside from Tinybird, all of the other managed database services would require building and maintaining at least one separate service to handle things like ETL and backend APIs.

The decision was simple.

The importance of iteration speed

As a team focused on improving CI flows for developers, Blacksmith's founders knew a thing or two about developer productivity and continuous iteration and deployment.

Rather than try to build the perfect analytics service over a protracted development cycle, Blacksmith wanted to ship fast. The goal was to get something in front of users that they could demonstrate, evaluate, and improve. They wanted to arrive at a rich and performant analytics offering as quickly as possible.

Tinybird gave them exactly the platform they needed to build quickly and test new features.

Abstracting complex data infrastructure as APIs

The engineers at Blacksmith use Tinybird's Events API, an HTTP endpoint that accepts JSON event payloads and handles writing them to the underlying database. They set up a pipeline to receive webhook events from customer jobs on GitHub, massage the payload to their spec, and forward that payload to the Events API.

Tinybird handles write throughput at up to 1,000 requests per second, and Blacksmith has used a micro-batching approach to easily scale writes into Tinybird.

Once the data lands in Tinybird, the team at Blacksmith uses Tinybird Pipes to build and iterate analytical queries to provide their end users with relevant metrics. Once they were happy with the data they received, they used Tinybird to publish and host API Endpoints to populate React dashboards on their Vercel-hosted application.

From zero to production in less than 3 weeks

Once the Blacksmith team settled on Tinybird as their analytics platform, they moved quickly and shipped their first user-facing workflow analytics experience in about two and a half weeks.

Finding new use cases

Blacksmith's founders initially chose Tinybird for user-facing analytics. But once they got data into Tinybird, they quickly discovered new use cases, specifically internal analytics and usage-based billing.

They use Tinybird for their own operational intelligence, using the data in Tinybird to build internal benchmarking systems to evaluate the performance of their infrastructure. They run nightly benchmark tests and dump that data into Tinybird, where they've built APIs that power dashboards showing them infrastructure performance over time.

They've also expanded the horizons of their usage-based billing system as enterprise customers have demanded custom pricing structures and discounts.

Moving faster for less

Speed is important, but cash is king. Blacksmith needed cost performance in their analytics infrastructure to continue to scale and grow the service.

Tinybird proved not only cost-effective but considerably less expensive than alternatives.

Blacksmith has just scratched the surface of their user-facing analytics experience, but they'll continue to dive deeper while relying on Tinybird to handle all of the infrastructure for their analytics and provide the perfect level of abstraction for their team.

With Tinybird, they'll continue focusing on their users and their analytics needs, rather than maintaining infrastructure.