Feb 16, 2024

How to scale a real-time data platform

Tinybird is an enterprise-grade data platform with large customers processing huge amounts of data. Learn how we scale to support their use cases.
Javier Santana
Co-founder

People often ask me how Tinybird will scale to support their use cases. It's a major source of anxiety; a lot of people have had bad experiences with vendors that claimed they could scale, throwing out buzzwords like "serverless" or "elastic" without really understanding the consequences of petabyte data loads and massive real-time architectures.

Tinybird is an enterprise-grade data platform that supports many very large customers processing very large amounts of data. We think about scale every day, every hour, every minute. We know how and when to scale infrastructure for our customers during peaks like Black Friday or when they start to experience steady-state growth. 

Scale is always on our minds. So, I thought I'd write about it.

Tinybird gives us everything we need from a real-time data analytics platform: security, scale, performance, stability, reliability, and a raft of integrations.

- Damian Grech, Director of Engineering, Data Platform at FanDuel

This is how Tinybird scales.

First and foremost, our philosophy for scaling is simple: "First, optimization. Then, infrastructure." Hardware is expensive, and logic is "free". So we always look for ways to optimize SQL queries to reduce the amount of resources required to serve their use cases.

But of course, it's more nuanced than that. To understand how Tinybird scales on a case-by-case basis, you must first understand the architecture. 

The basic Tinybird architecture

At a high level, Tinybird is a real-time data platform built around a real-time database (ClickHouse). We have several database replicas to manage our customers' loads, and the number of replicas varies based on what we're supporting at any given moment. The database replicas have shared storage, typically on S3 or GCS, and each replica has a fast cache in local SSD disks to speed up critical queries.

In front of the database, we host managed connectors for data ingestion from various streaming (e.g. Kafka, Confluent) and batch (e.g. Snowflake, S3) data sources, and we also manage things like user interactions, version control, query management, etc. 

All of these are integral pieces of the platform. We're not just a DBaaS scaling a database cluster for our customers. We're also scaling their streaming data ingestion and the high-concurrency, low-latency APIs they create from their database queries.

A simplified representation of the Tinybird architecture.

What we mean by "scale"

Also, to understand how Tinybird scales, you must understand what "scale" means. What order of magnitude are we talking about?

Here are some real-world examples from some of our customers that help to wrap your mind around our level of scale:

  • A top-five global clothing retailer uses Tinybird to build real-time product recommendations on their eCommerce site. On Black Friday in 2023, they ingested 4.3 billion events (7.5TB) of streaming data and supported 9.5k peak API requests per second with these latency metrics:
    • p90 - 56 ms
    • p99 - 139 ms
    • p999 - 247 ms
    • p9999  - 679 ms
    • Error rate: 0.000002%
  • A CDN client ingests into Tinybird all requests made to their system. This amounts to about ~250K events per second on average, with peaks of ~600K events per second.
  • Other customers build use cases with streaming data, such as real-time stock management, web analytics, internal analytics, and more. They typically load about 20-30K events per second from Apache Kafka or other streaming sources into Tinybird, with peaks of 300K events per second. 
  • Some customers ingest more than 1 million events per second.

So, roughly speaking, we're dealing with single customers that can exceed 1M events per second of streaming data ingestion, 10k API requests per second processing several petabytes of data every day, and achieving request latencies below 50 milliseconds for the majority of queries.

This is no small feat.

We had noticed how fast Tinybird could process events, so we decided to go even further and send more events from even more forms to Tinybird. We didn’t even have to warn them. The entire time, P99 latency stayed beneath 100ms.

- Juan Vega, Software Engineer at Typeform

There's no one way to scale

Of course, the scale depends on the use case. Some clients don't have large ingestion loads, but they make a lot of heavy concurrent queries. Some clients don't have big queries, but they need to have heavy loads of streaming data available for query in less than a second. Some customers have both.

A real-time analytics platform like Tinybird must account for customer scale as it relates to query latency, query complexity, access concurrency, data freshness, and historical data storage.

In general, here's how we scale to handle certain use cases:

Scaling query concurrency

Customers like FanDuel, Canva, Vercel, and many others build on Tinybird because they can create real-time APIs that serve thousands of concurrent requests in milliseconds. As customers scale their query concurrency, here's what we do to maintain latency below acceptable levels:

  • Add more replicas, OR
  • Add more CPUs to each replica, OR
  • Both

In this case, storage is common and doesn't necessarily need to scale. Adding replicas is straightforward and scales the compute necessary to achieve high concurrency at low latency.

Reducing latency

Our Enterprise customers often have strict latency SLAs that we're contractually obligated to keep. We are bound to maintain percentile response times below certain levels even as streaming ingestion increases or the customer experiences peak concurrency.

Often this looks like scaling individual queries. For critical use cases, we identify queries that can benefit from additional CPU cores, and scale them up. 

Of course, depending on the query, the way we scale can change. Sometimes it's as simple as changing a configuration parameter in the endpoint. Other times, we might use all the available replicas to run a single query in parallel.

The goal is always to ensure that all of our customer use cases, from top to bottom, can scale on demand.

Scaling ingestion

Remember, we must be able to simultaneously scale writes and reads, constantly aware of the impacts of ingestion on consumption and vice versa.

When it comes to scaling ingestion, we have a few tricks up our sleeve. Fortunately, we can use all of the replicas as writers if we need to because we use shared storage for the replicas, and any replica can write to it. 

That said, we're currently only relying on one writer in production. A lot of the scaling happens even before data hits the database. We have built and are constantly improving the services that handle ingestion, doing the heavy lifting before data touches the database.

Hardware is expensive, logic is free.

From a technical standpoint, we can always assign more CPUs or create more replicas, but scaling hardware is expensive. So, as I mentioned above, before we touch the infrastructure, we look at the software and the logic. 

Working in real-time requires a different mindset, especially a different way of thinking about constructing your SQL queries most efficiently. We support our customers by identifying underperforming or resource-hogging queries, and we work alongside them to optimize their queries to read less data.

Materialized Views are a powerful lever here for scaling without hardware. In Tinybird, Materialized Views can pre-calculate aggregates incrementally, as new data is ingested. They have the power to reduce query scan sizes by several orders of magnitude, so we use them often as optimization techniques for scaling concurrency and latency without adding hardware.

Beyond that, we help our customers follow best practices for writing SQL in real-time data pipelines so that they scan less data.

Some more reading about how we scale

We've written some great blog posts about scale in various aspects of our platform. I've consolidated them below for further reading in case you're interested:

Do you like this post?

Related posts

More Data, More Apps: Improving data ingestion in Tinybird
The realtime data stack. What’s next for Tinybird.
Tinybird has joined the AWS ISV Accelerate Program
Tinybird
Team
Jun 05, 2024
Tinybird is out of beta and open to everyone
Tinybird connects with Confluent for real-time streaming analytics at scale
Tinybird
Team
Jul 18, 2023
Building an enterprise-grade real-time analytics platform
Automating customer usage alerts with Tinybird and Make
$30M to lead the shift to real-time data
Tinybird expands self-service real-time analytics to AWS
Tinybird vs. ClickHouse®️: What's the difference?

Build fast data products, faster.

Try Tinybird and bring your data sources together and enable engineers to build with data in minutes. No credit card required, free to get started.
Need more? Contact sales for Enterprise support.