The Netflix TechBlog

MARCH 28, 2025

By Ko-Jen Hsiao , Yesu Feng and Sudarshan Lamkhede Motivation Netflixs personalized recommender system is a complex system, boasting a variety of specialized machine learned models each catering to distinct needs including Continue Watching and Todays Top Picks for You. Refer to our recent overview for more details).

Tuning 200

Tuning Efficiency Latency Strategy 200

Dynatrace

APRIL 10, 2025

Key insights for executives: Increase operational efficiency with automation and AI to foster seamless collaboration : With AI and automated workflows, teams work from shared data, automate repetitive tasks, and accelerate resolutionfocusing more on business outcomes. No delays and overhead of reindexing and rehydration.

Strategy 296

Strategy Storage Network Architecture 296

John McCalpin

FEBRUARY 17, 2025

Understanding sustained memory bandwidth in these systems starts with assuming 100% utilization and then reviewing the factors that get in the way (e.g., In my previous post , I reviewed historical data on single-core/single-thread memory bandwidth in multicore processors from Intel and AMD from 2010 to the present.

Latency 71

Latency Hardware Cache Systems 71

The Netflix TechBlog

SEPTEMBER 18, 2024

Second, developers had to constantly re-learn new data modeling practices and common yet critical data access patterns. These include challenges with tail latency and idempotency, managing “wide” partitions with many rows, handling single large “fat” columns, and slow response pagination.

Latency 260

Latency Storage Cache Efficiency 260

Scalegrid

FEBRUARY 6, 2025

RabbitMQ is designed for flexible routing and message reliability, while Kafka handles high-throughput event streaming and real-time data processing. Both serve distinct purposes, from managing message queues to ingesting large data volumes.

Latency 147

Latency Analytics Architecture Storage 147

The Netflix TechBlog

FEBRUARY 14, 2025

Every image you hover over isnt just a visual placeholder; its a critical data point that fuels our sophisticated personalization engine. It requires a state-of-the-art system that can track and process these impressions while maintaining a detailed history of each profiles exposure.

Tuning 166

Tuning Latency Efficiency Storage 166

Dynatrace

AUGUST 21, 2024

These releases often assumed ideal conditions such as zero latency, infinite bandwidth, and no network loss, as highlighted in Peter Deutsch’s eight fallacies of distributed systems. With Dynatrace, teams can seamlessly monitor the entire system, including network switches, database storage, and third-party dependencies.

Engineering 262

Engineering Systems Latency Metrics 262

Dynatrace

MAY 23, 2024

The jobs executing such workloads are usually required to operate indefinitely on unbounded streams of continuous data and exhibit heterogeneous modes of failure as they run over long periods. Fault tolerance stands as a critical requirement for continuously operating production systems. This significantly increases event latency.

Engineering 245

Engineering Tuning Latency Open Source 245

The Netflix TechBlog

OCTOBER 8, 2024

Rajiv Shringi Vinay Chella Kaidan Fullerton Oleksii Tkachuk Joey Lynch Introduction As Netflix continues to expand and diversify into various sectors like Video on Demand and Gaming , the ability to ingest and store vast amounts of temporal data — often reaching petabytes — with millisecond access latency has become increasingly vital.

Latency 239

Latency Storage Traffic Tuning 239

The Netflix TechBlog

OCTOBER 27, 2020

By Tianlong Chen and Ioannis Papapanagiotou Netflix has more than 195 million subscribers that generate petabytes of data everyday. Data scientists and engineers collect this data from our subscribers and videos, and implement data analytics models to discover customer behaviour with the goal of maximizing user joy.

Latency 252

Latency Storage Big Data Tuning 252

The Netflix TechBlog

MARCH 7, 2024

Berg , Romain Cledat , Kayla Seeley , Shashank Srikanth , Chaoying Wang , Darin Yu Netflix uses data science and machine learning across all facets of the company, powering a wide range of business applications from our internal infrastructure and content demand modeling to media understanding.

Systems 235

Systems Media Cache Open Source 235

The Netflix TechBlog

JULY 26, 2021

This happens at an unprecedented scale and introduces many interesting challenges; one of the challenges is how to provide visibility of Studio data across multiple phases and systems to facilitate operational excellence and empower decision making.

Big Data 257

Big Data Government Processing Analytics 257

The Netflix TechBlog

MARCH 7, 2023

These media focused machine learning algorithms as well as other teams generate a lot of data from the media files, which we described in our previous blog , are stored as annotations in Marken. But we cannot search or present low latency retrievals from files Etc. We do that by excluding the following from all queries in our system.

Media 272

Media Latency Architecture Database 272

The Netflix TechBlog

DECEMBER 17, 2024

To achieve this, we are committed to building robust systems that deliver comprehensive observability, enabling us to take full accountability for every title on ourservice. Each title represents countless hours of effort and creativity, and our systems need to honor that uniqueness. Yet, these pages couldnt be more different.

Traffic 172

Traffic Scalability Strategy Monitoring 172

Dynatrace

AUGUST 20, 2024

Driven by that value, Dynatrace brings real-time observability, security, and business data into context and makes sense of it so our customers can get answers, automate, predict, and prevent. Executives are sitting on a goldmine of data, and they don’t know it. Common business analytics incur too much latency.

Analytics 285

Analytics Latency Processing Systems 285

Dynatrace

JUNE 25, 2020

The network latency between cluster nodes should be around 10 ms or less. With Dynatrace actively managing business-critical applications, some of our globally distributed enterprise customers require Dynatrace Managed to continue operating even when an entire data center goes down. Minimized cross-data center network traffic.

Availability 264

Availability Hardware Latency Traffic 264

Dynatrace

OCTOBER 3, 2024

Considering the latest State of Observability 2024 report, it’s evident that multicloud environments not only come with an explosion of data beyond humans’ ability to manage it. It’s increasingly difficult to ingest, manage, store, and sort through this amount of data. You can find the list of use cases here.

Performance 290

Performance Architecture Innovation Latency 290

Dynatrace

SEPTEMBER 23, 2020

Recent improvements in OneAgent runtime-data handling. Operating Systems are not always set up in the same way. Storage mount points in a system might be larger or smaller, local or remote, with high or low latency, and various speeds. Customizable location of large runtime files. See details below. See details below.

Storage 227

Storage Latency Operating System Network 227

Scalegrid

FEBRUARY 24, 2025

Scaling RabbitMQ ensures your system can handle growing traffic and maintain high performance. Youll also learn strategies for maintaining data safety and managing node failures so your RabbitMQ setup is always up to the task. Monitoring the cluster nodes preemptively addresses potential issues, ensuring the system operates smoothly.

Best Practices 147

Best Practices Traffic Strategy Efficiency 147

The Netflix TechBlog

MAY 4, 2023

Behind the scenes, a myriad of systems and services are involved in orchestrating the product experience. These backend systems are consistently being evolved and optimized to meet and exceed customer and product expectations. It provides a good read on the availability and latency ranges under different production conditions.

Traffic 347

Traffic Latency Tuning Systems 347

Dynatrace

JANUARY 31, 2024

I have ingested important custom data into Dynatrace, critical to running my applications and making accurate business decisions… but can I trust the accuracy and reliability?” ” Welcome to the world of data observability. At its core, data observability is about ensuring the availability, reliability, and quality of data.

DevOps 272

DevOps Analytics Airlines Metrics 272

The Netflix TechBlog

DECEMBER 21, 2020

By Anupom Syam Background At Netflix, our current data warehouse contains hundreds of Petabytes of data stored in AWS S3 , and each day we ingest and create additional Petabytes. We built AutoOptimize to efficiently and transparently optimize the data and metadata storage layout while maximizing their cost and performance benefits.

Storage 212

Storage Latency Efficiency Data Engineering 212

DZone

FEBRUARY 5, 2024

In the fast-paced digital world, where every millisecond counts, understanding the nuances of network latency becomes paramount for developers and system architects. Latency, the delay before a transfer of data begins following an instruction for its transfer, can significantly impact user experience and system performance.

Latency 147

Latency Metrics Network Systems 147

Dynatrace

AUGUST 27, 2019

Step 2 – xMatters passes Dynatrace data to alerts that provide actionable responses. In this alert, xMatters includes all the important incident information from Dynatrace, so there’s no need for you to visit additional system dashboards. Dynatrace data in an xMatters alert (left), with actionable responses (right).

Systems 208

Systems DevOps Latency Azure 208

The Netflix TechBlog

SEPTEMBER 10, 2024

On Titus , our multi-tenant compute platform, a "noisy neighbor" refers to a container or system service that heavily utilizes the server's resources, causing performance degradation in adjacent containers. To emit a run queue latency metric, we leveraged three eBPF hooks: sched_wakeup, sched_wakeup_new, and sched_switch.

Latency 262

Latency Metrics Programming Monitoring 262

Dynatrace

APRIL 10, 2025

In modern containerized environments, teams often deploy Kubernetes across mixed operating systems, creating a situation where both Linux and Windows nodes reside in the same cluster. Integrating data at an OS-agnostic cluster level is another hurdle, often leading to data silos and incomplete visibility.

Performance 147

Performance Java Operating System Infrastructure 147

CSS Wizardry

JANUARY 31, 2019

When it comes to network performance, there are two main limiting factors that will slow you down: bandwidth and latency. the maximum rate of data transfer across a given path. Latency is defined as…. how long it takes for a bit of data to travel across the network from one node or endpoint to another. The Time Column.

Latency 137

Latency Network Speed Servers 137

Dynatrace

MAY 17, 2023

OpenTelemetry , the open source observability tool, has become the go-to standard for instrumenting custom applications to collect observability telemetry data. For this third and final part of our series, we saved the best for last: How you can enhance telemetry data even more and with less effort on your end with Dynatrace OneAgent.

Metrics 243

Metrics Database Monitoring Network 243

The Netflix TechBlog

JUNE 14, 2023

Testing Strategies: A Summary Two key factors determined our testing strategies: Functional vs. non-functional requirements Idempotency If we were testing functional requirements like data accuracy, and if the request was idempotent , we relied on Replay Testing. In such cases, we were not testing for response data but overall behavior.

Traffic 357

Traffic Latency Metrics Cache 357

Dynatrace

JULY 22, 2024

Using OpenTelemetry, developers can collect and process telemetry data from applications, services, and systems. Observability Observability is the ability to determine a system’s health by analyzing the data it generates, such as logs, metrics, and traces. There are three main types of telemetry data: Metrics.

Latency 264

Latency Best Practices Metrics Open Source 264

O'Reilly

MARCH 25, 2025

The system is inconsistent, slow, hallucinatingand that amazing demo starts collecting digital dust. Two big things: They bring the messiness of the real world into your system through unstructured data. People have been building data products and machine learning products for the past couple of decades. The way out?

Systems 69

Systems Development Tuning Monitoring 69

The Netflix TechBlog

NOVEMBER 3, 2022

As the number of Titus users increased over the years, the load and pressure on the system increased substantially. We introduce a caching mechanism in the API gateway layer, allowing us to offload processing from singleton leader elected controllers without giving up strict data consistency and guarantees clients observe.

Cache 233

Cache Latency Traffic Systems 233

Dynatrace

DECEMBER 27, 2022

SLOs can be a great way for DevOps and infrastructure teams to use data and performance expectations to make decisions, such as whether to release and where engineers should focus their time. This telemetry data serves as the basis for establishing meaningful SLOs. SLOs aid decision making. SLOs promote automation. Reliability.

Software 306

Software Software Benchmarking Latency 306

VoltDB

DECEMBER 11, 2024

Edge computing has transformed how businesses and industries process and manage data. By bringing computation closer to the data source, edge-based deployments reduce latency, enhance real-time capabilities, and optimize network bandwidth. Data interception during transit. Redundancy and inefficiency in data aggregation.

IoT 52

IoT Energy Logistics Latency 52

The Netflix TechBlog

DECEMBER 17, 2019

Andreas Andreakis , Ioannis Papapanagiotou Overview Change-Data-Capture (CDC) allows capturing committed changes from a database in real-time and propagating those changes to downstream consumers [1][2]. Requirements In a previous blog post, we discussed Delta , a data enrichment and synchronization platform.

Database 221

Database Traffic Transportation Open Source 221

High Scalability

JANUARY 11, 2022

from a client it performs two parallel operations: i) persisting the action in the data store ii) publish the action in a streaming data store for a pub-sub model. User Feed Service, Media Counter Service) read the actions from the streaming data store and performs their specific tasks. System Components. Data Models.

Design 334

Design Media Storage Logistics 334

Dynatrace

SEPTEMBER 29, 2020

Sydney, we have a disk write latency problem! It was on August 25 th at 14:00 when Davis initially alerted on a disk write latency issues to Elastic File System (EFS) on one of our EC2 instances in AWS’s Sydney Data Center. The problem didn’t last long or have any impact on our services.

Infrastructure 259

Infrastructure Cloud Monitoring AWS 259