How Adobe uses OpenTelemetry Collector (2023)

by AdobeChrisa Featherstone'aIShubhanshu Suranapraised the OpenTelemetry collector as the Swiss army knife of visibility in his speech onNorth America's Open Source Summit.

They went on to explain how they use it to track the massive amounts of viewability data their company collects, including metrics of 330 million unique rows per day. data range of 3.6 terabytes per day. and more than 1 petabyte of log data per day.

How Adobe uses OpenTelemetry Collector (1)

Adobeov Chris Featherstone and Shubhanshu Surana

Featherstone, senior director of software development, explained that not all of this data goes through his team or the Otel collector, but "it's a pretty good part of it."

Distributed tracking led his team toOpen telemetry. He explained that Adobe is mostly made up of acquisitions, and with each new company, people have their own opinions about the best cloud, this tool, this word processor, and so on.

"Especially with distributed sensors, this becomes a major challenge," he said. “Imagine trying to hurt an open source cloud provider. This finally led us to the collector. But we tried to build a distributed discovery platform based on Jaeger agents." It was in 2019.

Otel Collector was released in April 2020 to replace the Jaeger Agents. At the beginning, the collector was only supposed to absorb traces, but in September 2021 he brought measurements and logs are also trying to bring.

(Video) Getting started with the OpenTelemetry Collector

How Adobe uses OpenTelemetry Collector (2)

The team manages applications using open telemetry libraries, mostly automated instruments and mostly Java. Something enriches the app, it fetches Adobe-specific data, and enriches its pipelines as data flows into the collector. It has some custom extensions and editors, the team makes customizations with GitOps where possible.

“The aggregator is very dynamic and expands to multiple destinations with a single dataset, which has been huge for us. …We sometimes transfer data about collectors to other collectors for further processing. So it's the Swiss army knife of visibility," said Featherstone.

See more Adobe, delight developer productivity with a map that helps developers write better code faster.

ForJavaservices specifically have a base container and "if you are using a java image you should use it... It already has a lot of quality of life features built in including the Java OpenTelemetry facility in the container [The configuration is] taken from our documentation and is exactly how we configure it for Java.

“So we set the Jaeger endpoint to a local DaemonSet collector. We set the metrics exporter to Prometheus, set the multipliers, set additional resource attributes, set tracer, exporter to Jaeger. And we set the tracking picker to always off depending on the parent," he said, noting that all of this is built into the Java image.

So with these configurations, any running javaAs Kubernetesat Adobe is already involved in discovery. Everything configured this way goes through the collector.

"So everyone participates in the discovery by just turning around," he said. “Measurements, we tried to reduce friction, people will still have to extract these metrics from this exporter somehow. We've made it easy enough, but it's not automatic." He said about 75% of what they run is Java, but they are testing the same ideaNode.jsIPythonand other photos.

Data management

"They're doing a lot of enrichment, and they're also making sure they're not being sent covertly as part of our data or discovery metrics," said Surana, a research engineer at the company.

It uses multiple processes, including a reduce processor, as well as a custom processor in OpenTelemetry Collector that allows them to eliminate some fields they don't want to transmit in the background, which may be personal or other sensitive data. They are also used to enrich the data, as adding more fields such as service IDs, Kubernetes clusters, and regions helps narrow down the search.

(Video) The OpenTelemetry Collector | A Complete 1 Hour Workshop

“Adobe relies on active acquisitions and we launch many different products in different ecosystems. There's a high probability of conflicting service names across products or with similar microservice names, so we wanted to make sure that didn't happen," he said.

It also uses Adobe's service registry, where each service has a unique ID associated with the service name. Allows each Adobe engineer to uniquely identify the service in a unified discovery background.

"It also allows engineers to quickly look up things, even if they don't know the service or who owns it, they can look at our service record, find the engineer contact for that product or team, and call to resolve the issue," Surana said.

How Adobe uses OpenTelemetry Collector (3)

They also send data to multiple export destinations.

"This is probably the most common use case," he said. “Before OpenTelemetry Collector, Adobe engineering teams used different processes, different libraries in different formats. They sent it to vendor products, open source projects, and it was very difficult for us to get the engineering teams to change their backend or just make minor changes to the backend code or application code because the engineers have their own products and product requirements that they are working on.

“With the introduction of OpenTelemetry Collector, as well as the OTLP [OpenTelemetry Protocol] format, it has made our job a lot easier. we are able to send their data to many vendors, many tools with just a few changes on our side.”

Last year, they were able to send indexing data to three different backends simultaneously to test one specific engineering use case.

They now send the data to another set of Otel collectors at the boundary, where they can perform transformations such as inverse sampling, rule-based sampling, and performance-based sampling.

He said they are always looking for other ways to get richer information while sending less data in the background.

(Video) Automatic instrumentation with the OpenTelemetry Agent

"All this configuration is controlled by git. We use OpenTelemetry OperatorHelm diagrams primarily for our infrastructure use case. ... It removes the responsibility of engineers from being experts in a given field ... and makes setup very easy," he said.

Instrumentation automation with OpenTelemetry Operator allows engineers to pass just a few notes to automatically arrange their services for all different signals without writing a single line of code.

"This is huge for us," he said. It takes developer productivity to the next level."

How Adobe uses OpenTelemetry Collector (4)

They also created a custom extension on OpenTelemetry Collector using a custom authentication interface. They had two main requirements for this authentication system: the ability to use one system to securely send data to different backends, and the ability to protect both open source and vendor tools.

OpenTelemetry Collector includes a rich set of processes for creating data processes, including an attribute editor that allows you to add attributes to log data and matrix data. It allows you to transform, enrich or modify the transmitted data without the need for any action by application engineers. Adobe also uses it to improve the searchability of its wallpapers.

The memory throttling processor ensures that Otel never runs out of memory and controls the amount of space needed to maintain state. They also use the Range to Matrix Editor and the Service Graph Editor to generate trace data and create metric dashboards on the fly.

What's next?

According to Featherstone, two things: improving data quality, which means getting rid of data that no one will see, and limiting speed limits to the limit.

The collector allows the edge device to create rules and drop some data.

“For the measurements, imagine that we managed to assemble the collector itself. You know, maybe we don't need a 15-second detail, let's cut it down to five minutes and then send it," Featherstone said.

(Video) OpenTelemetry Course - Understand Software Performance

“Someone else might be sending some metrics for long-term storage and some for further processing in some operational data lake or something. We have the ability to rotate the manifold properly and do all sorts of things.”

Another thing is the speed limiter holes on the edge.

“We have one of our endpoints that gets around 60 billion visits a day and we're trying to keep track of that. This becomes a lot of data when you're talking about sending it somewhere for storage. So we are trying to figure out where are the right places to apply the rate limits to which collectors and at what levels…

They also try to focus more on solving problems first.

"We have so many east/west services that we're trying to do it through records and we're trying to get the right index of records for any group, and I have access to that or whatever. It's so slow and so hard to do that we're trying to really change the way people solve problems at Adobe to something like this where we've put a lot of effort into making these paths quite comprehensive," he said.

They also watch how people deal with problems and provide tools that have the best way to do it.

They are looking forward to integrating the OpenTelemetry logging libraries with core application libraries and running Otel collectors as sidebars for sending measurements, traces and records. They explore newconnectorand creating an edge tracking sampling extension to improve data quality.

Finally, he praised the architecture of the collector plugins and the ability to send data to different destinations with a single binary. That said, there is a rich set of extensions and editors that provide a lot of flexibility in data processing.

"Overall, OpenTelemetry feels very similar to the early days of Kubernetes, when everyone was buzzing a bit about it, and it started out like we were in a hockey rink right now," he said. “The community is great. The work is amazing. If you haven't messed with a collector yet, you should definitely check it out."

(Video) Deploying the OpenTelemetry Collector on Kubernetes | OpenTelemetry Fundamentals LIVE

Susan Hall is co-editor of The New Stack. Its task is to help sponsors reach as many recipients of their content as possible. She has written for The New Stack since its inception, as well as for the websites... Read more Susan Hall


How Adobe uses OpenTelemetry Collector? ›

OpenTelemetry Collector comes with a rich set of processes for building data processes, including an attribute processor which allows you to add attributes on top of log data and matrix data. It allows you to transform, enrich or modify the data in transit without the application engineers doing anything.

How does OpenTelemetry collector work? ›

OpenTelemetry Collector works by receiving telemetry data from various sources, processing and normalizing the data, and then exporting it to various backends for storage and analysis.

What is the difference between OpenTelemetry agent and collector? ›

OpenTelemetry: agent vs collector vs gateway

Basically, an agent collects the data from the host and sends it to the collector to process and export. The agent is an OpenTelemetry Collector instance running on the same host as the application you're collecting the telemetry data from.

What is collector in open telemetry? ›

OpenTelemetry (abbreviated "OTel") is a shared effort between the community and observability vendors to establish a common specification for telemetry in software systems: traces, metrics, and logs. The OTel Collector is an application that allows you to process that telemetry and send it out to various destinations.

What are the components of the OTel collector? ›

The OTEL Collector Architecture. The OpenTelemetry collector consists of three main components: receivers, processors, and exporters, which are then used to construct telemetry pipelines.

What are the benefits of OpenTelemetry collector? ›

A few key benefits of OpenTelemetry are:
  • Vendor Neutrality. With OpenTelemetry, you can collect telemetry data from different sources and send it to multiple platforms without significant configuration changes. ...
  • Data Flexibility. OpenTelemetry allows you to control what telemetry data you send to your platforms. ...
  • Easy Setup.

What are the advantages of OpenTelemetry collector? ›

The collector provides a vendor-neutral way to collect, process, and export your telemetry data(logs, metrics, and traces), and that's why it is preferable to use a collector. The biggest advantage of using OpenTelemetry collectors is the flexibility to create different data pipelines.

What is the difference between OpenTelemetry and app metrics? ›

Key differences between OpenMetrics and OpenTelemetry​

OpenMetrics just standardizes the exposition format through which metrics data must be communicated over a network, whereas OpenTelemetry aims to provide compatibility with OpenMetrics by using this standardized format to export metrics data.

Where is OpenTelemetry used? ›

OpenTelemetry is used to collect telemetry data from distributed systems in order to troubleshoot, debug and manage applications and their host environment. It offers an easy way for IT and developer teams to instrument their code base for data collection and make adjustments as the organization grows.

What is the difference between OpenTelemetry API and SDK? ›

API allows to construct the Metric of a chosen type. SDK defines the way to query the current value of a Metric to be exported. Every type of a Metric has it's API to record values to be aggregated. API supports both - push and pull model of setting the Metric value.

What is the alternative to OpenTelemetry collector? ›

Top Alternatives to OpenTelemetry
  • Kibana. Kibana is an open source (Apache Licensed), browser based analytics and search. ...
  • Grafana. Grafana is a general purpose dashboard and graph composer. ...
  • Prometheus. Prometheus is a systems and service monitoring system. ...
  • Nagios. ...
  • Zabbix. ...
  • Graphite. ...
  • Jaeger. ...
  • StatsD.

What are the three types of telemetry? ›

Types of telemetry system are: Aerospace telemetry. Medical telemetry. Meteorological telemetry.

What are the three pillars of telemetry? ›

Peak visibility is achieved by analyzing the three pillars of observability: Logs, metrics and traces. Depending on who you ask, some use MELT as the four pillars of essential telemetry data (or metrics, events, logs and traces) but we'll stick with the three core pillars for this piece.

How do you collect logs in OpenTelemetry? ›

OpenTelemetry recommends to collect application logs using Collector's filelog receiver. Alternatively, another log collection agent, such as FluentBit, can collect logs, then send to OpenTelemetry Collector where the logs can be further processed and enriched.

What is OTel used for? ›

OpenTelemetry (also referred to as OTel) is an open source observability framework made up of a collection of tools, APIs, and SDKs. Otel enables IT teams to instrument, generate, collect, and export telemetry data for analysis and to understand software performance and behavior.

How do I set up OpenTelemetry? ›

  1. Prerequisites.
  2. Example Application.
  3. Installation. Create and launch an HTTP Server.
  4. Instrumentation.
  5. Run the instrumented app.
  6. Add manual instrumentation to automatic instrumentation. Traces. Metrics.
  7. Send telemetry to an OpenTelemetry Collector. Configure and run a local collector. ...
  8. Next steps.

What are the disadvantages of OpenTelemetry? ›

Disadvantages of OpenTelemetry​

Some of the key disadvantages of using OpenTelemetry are: Project has a lot of scope in improving documentation and support. It does not provide backend storage and a visualization layer.

What is the difference between OpenTelemetry and open tracing? ›

OpenTelemetry and OpenTracing are open-source projects used to instrument application code for generating telemetry data. While OpenTelemetry can help you generate logs, metrics, and traces, OpenTracing focuses on generating traces for distributed applications.

What is distributed tracing vs OpenTelemetry? ›

Differences Between OpenTelemetry and Distributed Tracing

One of the differences between the two approaches is that OpenTelemetry focuses on collecting data at the application level, while distributed tracing also includes data from infrastructure components like databases and message queues.

How does OpenTelemetry auto instrumentation work? ›

An application can be instrumented with OpenTelemetry either automatically or manually. With automatic instrumentation, the API and SDK takes the configuration provided (through code or environment variables) and does most of the work automatically (for example, starting and ending spans).

What does OTLP mean in OpenTelemetry? ›

The OpenTelemetry Protocol (OTLP) specification describes the encoding, transport, and delivery mechanism of telemetry data between telemetry sources, intermediate nodes such as collectors, and telemetry backends.

What is the difference between Prometheus and OpenTelemetry? ›

OpenTelemetry helps companies generate telemetry data by instrumenting their code. Prometheus, on the other hand, monitors metrics. There are client libraries for both Prometheus and OpenTelemetry, but OTEL provides an all-in-one solution for logging, metrics, and tracing your code. Prometheus only generates metrics.

What is difference between Jaeger and OpenTelemetry? ›

The key differences between OpenTelemetry and Jaeger can be summarized in the following points: OpenTelemetry provides an instrumentation layer for your application code, while Jaeger is a backend analysis tool used for storage and visualization of trace data.

Does Prometheus use open telemetry? ›

Prometheus exporter

Once the exporter is configured and enabled, it will collect metrics from the OpenTelemetry Collector and forward them to the specified Prometheus server. Users can then use Prometheus tools to query, visualize, and alert on the collected metrics.

Does Splunk support open telemetry? ›

We have built Splunk Observability on top of OpenTelemetry, so no matter your instrumentation or data collection agent, you can pull in your data with ease.

What is the difference between OpenTelemetry and DataDog? ›

OpenTelemetry and DataDog are both used for monitoring applications. While OpenTelemetry is an open source observability framework, DataDog is a cloud-monitoring SaaS service. OpenTelemetry is a collection of tools, APIs, and SDKs that help generate and collect telemetry data (logs, metrics, and traces).

Who owns open telemetry? ›

Michael Hausenblas is the OpenTelemetry product owner and Solution Engineering Lead in the AWS open source observability service team.

What are the 2 types of API accessibility? ›

A public API and an internal API, for example, are two distinct types, based on who has access.

Why use SDK instead of API? ›

APIs are used to communicate between different applications, whilst an SDK is a tool kit to build applications and features. In most cases, an SDK application will contain an API — or even multiple APIs depending on the purpose of the software.

What is SDK vs API data collection? ›

Difference between SDK and API

An SDK is basically a toolbox with everything third-party developers need to write applications for a specific platform or programming language. This toolbox typically includes at least one API. An API is an interface that enables different applications to work together.

How many types of telemetry are there? ›

There are four key telemetry data types: Metrics.

What is the difference between telemetry and IoT? ›

Applications using the technology include measuring and transmitting data from internet of things (IoT) sensors located in automobiles, smart meters, power sources, robots and even wildlife. Telemetry sends data using computer networks, satellites, cable, and infrared and ultrasonic technologies.

What is one of the most common information collected using telemetry? ›

Companies use telemetry to collect information about devices and their configuration. Such information includes storage, installed memory, and hardware components, such as the CPU.

What are the two types of telemetry monitoring? ›

Overview of Holter and Ambulatory Telemetry Monitors

There are two common types of ambulatory monitors — Holter monitors and ambulatory telemetry monitors (ATMs).

What are the four distinct types of telemetry data? ›

  • 3.1.1. Management Plane Telemetry. ...
  • 3.1.2. Control Plane Telemetry. ...
  • 3.1.3. Forwarding Plane Telemetry. ...
  • 3.1.4. External Data Telemetry.
Dec 3, 2021

How many leads are used for telemetry? ›

Prior to commencing telemetry monitoring a baseline 12 or 15 lead ECG should be taken as per cardiac team. Additional ECG's should be performed if ECG changes from patient baseline.

What types of signal transmissions are used in telemetry? ›

(d) Signal Transmission Medium: It is the medium or link that connects the sending or transmitting end to the receiving end, over which the transmitter can transmit its output signal to the receiver. Broadly, there are three signal transmission media in use: (i) copper wires, (ii) radio link, and (iii) optical fibre.

What format is OpenTelemetry tracing? ›

The default format used in OpenTelemetry tracing is W3C TraceContext . Each Context object is associated with a span and can be accessed specification on spans.

How do you test OpenTelemetry? ›

Send a Test Span Through an OpenTelemetry Collector
  1. Put a test span in a file.
  2. Send that data to the collector (using curl or Postman)
  3. Find the span in Honeycomb.
Dec 16, 2022

What format is open telemetry message? ›

OTLP/HTTP. OTLP/HTTP uses Protobuf payloads encoded either in binary format or in JSON format.

What is the size of the collector in otel? ›

By default, the Collector is configured to use 512 MB of memory.

How is telemetry data collected? ›

Telemetry automatically collects, transmits and measures data from remote sources, using sensors and other devices to collect data. It uses communication systems to transmit the data back to a central location. Subsequently, the data is analyzed to monitor and control the remote system.

What is the difference between logging and telemetry? ›

Logging is the local collection of data while telemetry is the remote transmission of data. While there is an overlap between the two topics, they serve different purposes, they have wildly different technical challenges and they must adhere to distinctly different regulatory requirements too.

What is the difference between telemetry and OpenTelemetry? ›

Telemetry data includes logs, metrics, and traces. It was formed after the merger of OpenTracing and OpenCensus, two projects with similar goals but different approaches. OpenTelemetry is a collection of APIs, SDKs, and client libraries used to generate telemetry data from your application code.

What are the three typical telemetry data is? ›

We use metrics, structured logs, and traces as the outputs to make our systems observable.

What problem does OpenTelemetry solve? ›

OpenTelemetry (also referred to as OTel) is an open source observability framework made up of a collection of tools, APIs, and SDKs. Otel enables IT teams to instrument, generate, collect, and export telemetry data for analysis and to understand software performance and behavior.

What is the difference between API and SDK in OpenTelemetry? ›

API allows to construct the Metric of a chosen type. SDK defines the way to query the current value of a Metric to be exported. Every type of a Metric has it's API to record values to be aggregated. API supports both - push and pull model of setting the Metric value.

What is Prometheus client vs OpenTelemetry? ›

OpenTelemetry helps companies generate telemetry data by instrumenting their code. Prometheus, on the other hand, monitors metrics. There are client libraries for both Prometheus and OpenTelemetry, but OTEL provides an all-in-one solution for logging, metrics, and tracing your code. Prometheus only generates metrics.

What is the difference between Jaeger client and OpenTelemetry? ›

The key differences between OpenTelemetry and Jaeger can be summarized in the following points: OpenTelemetry provides an instrumentation layer for your application code, while Jaeger is a backend analysis tool used for storage and visualization of trace data.


1. Controlling Data Overhead with the OpenTelemetry Collector - Nočnica Mellifera, TelemetryHub
(The Linux Foundation)
2. Distributed Tracing in .NET 6 using OpenTelemetry - Martin Thwaites - NDC Copenhagen 2022
(NDC Conferences)
3. Otel Collector: The Swiss Army Knife of Observability - Chris Featherstone & Shubhanshu Surana
(The Linux Foundation)
4. You MUST Instrument Your Code With OpenTelemetry!
(DevOps Toolkit)
5. OpenTelemetry Agent and Collector: Telemetry Built-in Into All S... Steve Flanders & Trask Stalnaker
(CNCF [Cloud Native Computing Foundation])
6. OpenTelemetry – An Observability Framework for Cloud Native Software by Dennis Kieselhorst


Top Articles
Latest Posts
Article information

Author: Reed Wilderman

Last Updated: 02/10/2023

Views: 6199

Rating: 4.1 / 5 (52 voted)

Reviews: 83% of readers found this page helpful

Author information

Name: Reed Wilderman

Birthday: 1992-06-14

Address: 998 Estell Village, Lake Oscarberg, SD 48713-6877

Phone: +21813267449721

Job: Technology Engineer

Hobby: Swimming, Do it yourself, Beekeeping, Lapidary, Cosplaying, Hiking, Graffiti

Introduction: My name is Reed Wilderman, I am a faithful, bright, lucky, adventurous, lively, rich, vast person who loves writing and wants to share my knowledge and understanding with you.