Open Graph Title

Introducing Faasly

The real-time
data platform for
observability

Why did we build FaaSly?

Most monitoring tools do too much

They install an agent or poll data through an API, send them to their time-series database, and then provide you with dashboards and charts to look at. This may seem good at first, but it causes more pain than cure when it comes time to scale up to terabytes per day!

Data visibility

Deciding what to log and what not to log and resorting to compression or sampling to keep the costs low.
Latency and throughput

Centralized logging has done more harm than good. It takes forever to scan the dataset and get the results.
Pricing model

Pricing plans that are stuck in the 90's. Vendors have exploited companies with GBs or rows ingested and retained.

How's Faasly different?

Our approach is simple: ingest the OpenTelemetry data in an S3 bucket as Parquet files in Iceberg table format and query them using DuckDB with milliseond retrieval and zero egress cost.

01. Run fast queries, faster ― State of the art DuckDB OLAP SQL query engine on cold S3 data.
02. Streaming data done right ― Kafka-compatible data streaming built directly on top of S3.
03. Bring your team on the same table ― Apache Iceberg brings a different table design for big data.
04. Fetch query results faster ― Parquet enables targeted searches requiring less I/O and avoid false positives.
05. Think you know scale? Think S3 scale ― 100M reqs/second. 400 Terabits/second. 280T objects.

Fork in the road

Let's talk about why we went with certain architectural choices

"This will never work". We've heard this a lot of times. We got a ton of advice to build this on top of ClickHouse. In fact, almost all modern day startups in this space use ClickHouse. A couple of issues that were raised in our choice of tech stack:

01. DuckDB is not built for large datasets ― 95% of queries on metrics data are for the last hour, and 98% of the time they are for the last 4 hours. Iceberg's time travel and version travel queries, along with DuckDB's parallel query processing and in-memory representation with Arrow, make it a great choice for handling larger than memory datasets.
02. DuckDB is not designed to run on multiple machines ― Two words - scale up and single tenancy. We love safely running our application on a single server and scale up as your data grows. We believe in scaling up rather than scaling out. Like our friends at Motherduck asks, why did we scale out in the first place?

Behind the screens

Scaleble, fault-tolerant and efficient data lakehouse

Our approach is opinionated, offering tools and systems for monitoring and security. This means that you do not have to build another observability pipeline, provision a Kafka cluster, or manage Spark jobs. We will take care of chaining everything for you.

OpenTelemetry

The gold standard in data collection

Instrument and ingest anything and everything as long as OpenTelemetry supports it. Run processors and export them to S3 without breaking a sweat.
Apache Flink

Stateful streaming analytics on cold S3

End-to-end pipeline with OpenTelemetry pushing data into Kafka, transforming it in Flink, and finally sinking everything into an Iceberg table as Parquet files in S3.
Apache Arrow

Efficient analytics on larger than memory datasets

The zero-copy integration between DuckDB and Apache Arrow allows for rapid analysis of larger than memory datasets in Python using SQL.
SQLAlchemy

Publish query results as low-latency APIs

Instantly publish query results as high-concurrency, low-latency REST APIs for your entire team to consume. Consistent results everytime.

Public Roadmap

Join our community of early adopters and advocates and help us prioritize

238K

Lines of code
218

Active Linear issues
88

Production deployments

- Log Management — done.
- Infrastructure Monitoring — done.
- Cloud SIEM — done.
- Distributed Tracing — in-progress.
- Email & Slack Alerts — in-progress.
- Cloud CSPM — in-progress.
- Continuous Profiler — planned.
- Cloud Service Map — planned.
- Cloud Asset Discovery — planned.
- Collaborative Workspace — done.
- Keyboard Shortcuts — done
- Custom Otel Agent — in-progress.
- Linear & GitHub Sync — in-progress
- Real User Monitoring — planned.
- Synthetic Monitoring — planned.

Ready to roll with us?

The realtime data platform
for engineering teams

Our mission is to make your work easier.
We make sure you never have to slog through logs again.

Join Waitlist →

l

Got any questions?

How's Faasly different?