---
title: "Honey, I shrunk the telemetry"
slug: "honey-i-shrunk-the-telemetry"
blurb: "We are excited to announce the release of bitdrift’s first product, Capture, as well as our $15M series A financing from an amazing group of investors led by Amplify Partners! Focused on mobile observability, we believe that Capture will revolutionize how mobile engineers debug their applications."
author:
  - "matt"
publishedDate: "2023-12-04T10:01:16.209Z"
modifiedDate: "2023-12-04T10:01:16.209Z"
tags:
  - "observability"
  - "mobile"
cover:
  url: "/assets/posts/honey-i-shrunk-the-telemetry/feature-blog-hero-image.svg-desktop@1x.webp"
  alt: "Honey, I shrunk the telemetry"
socialThumbnail:
  url: "/assets/posts/honey-i-shrunk-the-telemetry/feature-blog-hero-image.svg-desktop@1x.webp"
  alt: "Honey, I shrunk the telemetry"

---

Today we are incredibly excited to announce the release of bitdrift’s first product,
[Capture](https://bitdrift.io), as well as our $15M series A financing from an [amazing group of
investors](https://bitdrift.io/about#investors) led by [Amplify
Partners](https://www.amplifypartners.com/)! Focused on mobile observability, we believe that
Capture will revolutionize how mobile engineers debug their applications. To be bold: Capture
provides mobile developers the richest and most dynamic observability on the planet without breaking
the bank.

Before diving more into Capture, we want to step back and talk a bit about why we founded bitdrift
in the first place.

bitdrift was born out of years of frustration while building internet infrastructure at scale. With
collective experience (and tears!) at companies such as Twitter, AWS, Square, Google, Microsoft,
Netflix, and Salesforce, including creating [Envoy](https://www.envoyproxy.io/) during our shared
tenure at Lyft, we came to realize an ugly truth: the observability ecosystem as it exists today is
misaligned between vendor and consumer: vendors charge by volume with little thought given to the
usefulness of stored data; consumers are frustrated both by their bills and their ability to solve
real customer problems with that stored data. What if it were possible to change the observability
narrative by using real time dynamic control to emit only the telemetry that would likely be used to
solve customer problems? This is exactly what we set out to do several years ago at Lyft.

## Mobile observability today is wasteful, disorganized, and way behind server

As the industry has evolved, our server frameworks and tools have improved by leaps and bounds, and
today we have the best visibility we’ve ever had. But we’re drowning in a sea of noise: 95% of the
data collected to monitor the health of systems is never read. Not by an engineer, not by a machine,
not by a data scientist, not by anyone! Yet the vendors and tools they provide are geared solely
toward greater and greater ingestion, happy to charge us for the privilege of storing useless data.
It’s an egregiously offensive misuse of time, money, and carbon.

At the same time, mobile observability is decades behind what is available on server. Mobile
engineers are lucky to have a static set of analytic events in production, and modifying them to
debug ongoing issues is likely a multi-week or month process to deploy changes out to the majority
of clients. Even when deployed, great care must be taken to not send too much data lest it overwhelm
limited network and CPU bandwidth and balloon data warehouse costs.

And finally, on both mobile and server there’s a huge gap in understanding and fixing the problems
users of internet systems are actually having. A sampled error log is only so useful. What about all
of the debug context that came before that log that might provide the clues needed to actually solve
the problem? What about high volume telemetry that is simply infeasible to send at all times?

We have been trained to capture data, and store and analyze it centrally. What if it were possible
to flip this around such that in the majority of cases we capture, store, and perform the first
phase of analysis remotely, and only commit to central storage in the cases in which the data is
highly likely to be utilized? Can we provide the illusion of unlimited, free telemetry?

## We’ve built a way to dynamically control telemetry in real time

Starting with mobile and [Capture](https://bitdrift.io), we are changing the observability game by
enabling dynamic real time control of emitted session telemetry on both iOS and Android. Devices can
be targeted instantaneously, from all clients, to specific cohorts (all Android users, all iOS users
on a particular OS version, etc.), all the way down to *individual devices*. [Sophisticated local
storage](https://bitdrift.io/buffer) coupled with [real time
configuration](https://bitdrift.io/realtime) via the bitdrift control plane allows for *distributed
search* over observability data, and for telemetry to be flushed *only* when asked for and when it
is highly likely to be useful in solving a customer problem. Once flushed, session data can be
viewed in a purpose built timeline viewer, facilitating rapid debugging of customer problems.

![a diagram representing the flow of a workflow to a device](/assets/workflows@2x.png)

The backbone of our local storage solution is what we call the “ring buffer.” The ring buffer is a
highly performance tuned subsystem that has been designed to use a bounded and real time
configurable amount of RAM and disk. Data is first flushed to RAM and then cascaded to disk in the
background. The ring buffer abstraction allows for cheap local storage of telemetry, and is the
basis on which the bitdrift family of products is built.

![a visual representation of a ring buffer superimposed over a mobile device displaying the lyft mobile app](/assets/ring_buffer@2x.png)


Capture also includes a highly efficient and privacy conscious implementation of [session
replay](https://bitdrift.io/replay), capturing both 2D and 3D representations of mobile screen
state. Unlike competing solutions, the screen capture storage is so small that it can be
continuously captured, leading to a vastly improved mobile debugging experience.

![a screenshot of the 3d view of session replay](/assets/capture@2x.png)

## Battle tested and ready to solve real world problems

Possibly most importantly, today’s SDK release for iOS and Android is not beta quality. It is
already deployed on millions of devices within the Lyft app, and has been *battle tested at scale*.
Capture is ready to solve real-world challenges for organizations around the world today.

![a screenshot of an IDE using the sdk](/assets/sdk@2x.png)

We have spent a large portion of our careers being shackled and frustrated by lengthy mobile release
cycles leading to multi-week and multi-month delays fixing customer issues. Our goal with Capture is
for mobile developers to collect as much telemetry as desired in their apps, because it’s free.
*Telemetry will only be emitted when called for by dynamic real time control*. We believe this
capability is going to unlock rapid resolution of customer issues at an unheard of low price point.

Today we are thrilled to open the waitlist for Capture. [Sign up](https://bitdrift.io/signup) to
give the [free tier](https://bitdrift.io/pricing) a try; we will be onboarding people rapidly while
we monitor and ensure a great experience for all. Please join us in
[Slack](https://communityinviter.com/apps/bitdriftpublic/bitdrifters) as well to ask questions and
give feedback!

## The beginning of a journey

And finally, Capture is the beginning of a journey that we are very, very excited to share with the
industry. [Mobile is only the first step](https://bitdrift.io/server). Local telemetry storage
coupled with real time control and distributed search is broadly applicable to the entire
distributed system: from every server all the way to the mobile edge.

Over the coming months we are excited to share technical blog posts on the technology that makes our
products possible. Watch this space and reach out to
[careers@bitdrift.io](mailto:careers@bitdrift.io) if working on this sounds interesting; we aren’t
hiring immediately but things change quickly!

Our company ethos: allow our customers to dynamically turn the visibility dial way down when not
needed, way up when things are broken, thus ingesting the right data, at the right time, from the
right sources.

Welcome to the future of observability.