The Netflix TechBlog

JANUARY 10, 2024

This architecture shift greatly reduced the processing latency and increased system resiliency. We expanded pipeline support to serve our studio/content-development use cases, which had different latency and resiliency requirements as compared to the traditional streaming use case.

Processing 229

Processing Media Latency Innovation 229

Adrian Cockcroft

JANUARY 20, 2023

Here’s some predictions I’m making: Jack Dongarra’s efforts to highlight the low efficiency of the HPCG benchmark as an issue will influence the next generation of supercomputer architectures to optimize for sparse matrix computations. Next generation architectures will use CXL3.0 Next generation architectures will use CXL3.0

Architecture 52

Architecture Latency Benchmarking AWS 52

The Netflix TechBlog

MARCH 7, 2023

We designed a unique concept called Annotation Operations which allows teams to create data pipelines and easily write annotations without worrying about access patterns of their data from different applications. But we cannot search or present low latency retrievals from files Etc. This is obviously very expensive.

Media 264

Media Latency Architecture Database 264

The Netflix TechBlog

NOVEMBER 3, 2022

In the time since it was first presented as an advanced Mesos framework, Titus has transparently evolved from being built on top of Mesos to Kubernetes, handling an ever-increasing volume of containers. The original assumptions and architectural choices were no longer viable.

Cache 224

Cache Latency Traffic Systems 224

Dynatrace

JANUARY 31, 2024

While off-the-shelf models assist many organizations in initiating their journeys with generative AI (GenAI), scaling AI for enterprise use presents formidable challenges. Figure 1: Sample RAG architecture While this approach significantly improves the response quality of GenAI applications, it also introduces new challenges.

Cache 209

Cache Azure Infrastructure Monitoring 209

The Netflix TechBlog

AUGUST 29, 2023

Today we have a wealth of tools, both OSS and commercial, all designed for cloud-native environments. To improve availability, we designed systems where components could fail separately and avoid single points of failure. Eureka and Ribbon presented a simple but powerful interface, which made adopting them easy.

Traffic 220

Traffic Latency Cloud C++ 220

Scalegrid

MARCH 14, 2024

The architecture usually integrates several private, public, and on-premises infrastructures. Key Components of Hybrid Cloud Infrastructure A hybrid cloud architecture usually merges a public Infrastructure-as-a-Service (IaaS) platform with private computing assets and incorporates tools to manage these combined environments.

Strategy 130

Strategy Cloud Artificial Intelligence Infrastructure 130

Scalegrid

FEBRUARY 8, 2024

Their design emphasizes increasing availability by spreading out files among different nodes or servers — this approach significantly reduces risks associated with losing or corrupting data due to node failure. Variations within these storage systems are called distributed file systems.

Storage 130

Storage Systems Big Data Azure 130

The Netflix TechBlog

JUNE 13, 2023

Our previous blog post presented replay traffic testing — a crucial instrument in our toolkit that allows us to implement these transformations with precision and reliability. A/B testing is also a key technique in migrations where the updates to the architecture involve changing device contracts as well.

Traffic 279

Traffic Metrics Systems Strategy 279

The Netflix TechBlog

MARCH 1, 2021

It supports both high throughput services that consume hundreds of thousands of CPUs at a time, and latency-sensitive workloads where humans are waiting for the results of a computation. The subsystems all communicate with each other asynchronously via Timestone, a high-scale, low-latency priority queuing system.

Serverless 228

Serverless Media Latency Social Media 228

ACM Sigarch

MAY 31, 2023

Introduction Memory systems are evolving into heterogeneous and composable architectures. However, building and utilizing HCM presents challenges, including interconnecting various memory technologies (e.g., However, building and utilizing HCM presents challenges, including interconnecting various memory technologies (e.g.,

Latency 52

Latency Hardware Cache Architecture 52

Percona

NOVEMBER 3, 2023

Rather than listing the concepts, function calls, etc, available in Citus, which frankly is a bit boring, I’m going to explore scaling out a database system starting with a single host. And now, execute the benchmark: -- execute the following on the coordinator node pgbench -c 20 -j 3 -T 60 -P 3 pgbench The results are not pretty.

Database 102

Database Benchmarking Latency C++ 102

The Netflix TechBlog

FEBRUARY 9, 2021

Plus, the architecture of the Edge tier was evolving to a PaaS (platform as a service) model, and we had some tough decisions to make about how, and where, to handle identity token handling. The system architecture now takes the form of: Notice that tokens never traverse past the Edge gateway / EAS boundary. We are serving over 2.5

Architecture 221

Architecture Latency Servers Website 221

Percona

APRIL 17, 2023

The swap issue is explained in the excellent article by Jeremy Cole at the Swap Insanity and NUMA Architecture. The CFQ works well for many general use cases but lacks latency guarantees. The deadline excels at latency-sensitive use cases ( like databases ), and noop is closer to no schedule at all.

Best Practices 52

Best Practices Design Tuning Database 52

The Netflix TechBlog

FEBRUARY 9, 2021

In particular, it’s our job to design and build the systems and protocols that enable customers from all over the world to sign up for Netflix with the plan features and incentives that best suit their needs. When we merge these two concepts together and present them to the customer, we have the plan selection page (shown above).

Engineering 142

Engineering Scalability Architecture Innovation 142

The Netflix TechBlog

FEBRUARY 9, 2021

Before designing a solution it’s important to understand the main product requirements for such a feature: The content needs to be new, relevant, and regional (not all countries have the same catalogue). We need to be able to easily determine what imagery is present for a given platform, region, and language.

Engineering 165

Engineering Storage Latency Entertainment 165

The Netflix TechBlog

MAY 5, 2021

In the future posts, we will do an architectural deep dive into the several components of Netflix Drive. To support such use cases, access control at the user workspace and project workspace granularity is extremely important for presenting a globally consistent view of pertinent data to these artists.

Media 174

Media Storage Architecture Cloud 174

The Netflix TechBlog

OCTOBER 27, 2020

At Netflix, we also heavily embrace a microservice architecture that emphasizes separation of concerns. The data warehouse is not designed to serve point requests from microservices with low latency. Therefore, we must efficiently move data from the data warehouse to a global, low-latency and highly-reliable key-value store.

Latency 243

Latency Storage Big Data Tuning 243

The Netflix TechBlog

AUGUST 30, 2021

System Setup Architecture The following diagram summarizes the architecture description: Figure 1: Event-sourcing architecture of the Device Management Platform. A timeline of the transition from Spring KafkaListener to Alpakka-Kafka is presented here for a better understanding of the motivations for the transition.

Latency 213

Latency Traffic Transportation Hardware 213

The Netflix TechBlog

JUNE 4, 2019

Because microprocessors are so fast, computer architecture design has evolved towards adding various levels of caching between compute units and the main memory, in order to hide the latency of bringing the bits to the brains. This avoids thrashing caches too much for B and evens out the pressure on the L3 caches of the machine.

Cache 251

Cache Latency Airlines Logistics 251

The Netflix TechBlog

DECEMBER 21, 2020

This article will list some of the use cases of AutoOptimize, discuss the design principles that help enhance efficiency, and present the high-level architecture. These principles reduce resource usage by being more efficient and effective while lowering the end-to-end latency in data processing.

Storage 203

Storage Latency Efficiency Data Engineering 203

Brendan Gregg

FEBRUARY 28, 2023

My personal opinion is that I don't see a widespread need for more capacity given horizontal scaling and servers that can already exceed 1 Tbyte of DRAM; bandwidth is also helpful, but I'd be concerned about the increased latency for adding a hop to more memory. Ford, et al., “TCP

Performance 110

Performance Latency Cache Virtualization 110

The Netflix TechBlog

MARCH 4, 2024

In this blog post, we present our project on Auto Remediation, which integrates the currently used rule-based classifier with an ML service and aims to automatically remediate failed jobs without human intervention. Upon further profiling, we found that most of the latency came from the candidate generated step (i.e.,

Tuning 210

Tuning Efficiency Big Data Engineering 210

The Netflix TechBlog

MARCH 13, 2020

The teams have been working closely on SVT-AV1 development, discussing architectural decisions, implementing new tools, and improving compression efficiency. SVT-AV1 also includes extensive documentation on the encoder design targeted to facilitate the onboarding process for new developers.

Open Source 154

Open Source Efficiency C++ Speed 154

John McCalpin

APRIL 2, 2020

This was a keynote presentation at the “2nd International Workshop on Performance Modeling: Methods and Applications” (PMMA16), June 23, 2016, Frankfurt, Germany (in conjunction with ISC16 ). This data is from the 2007 presentation.

Benchmarking 59

Benchmarking Performance Latency Architecture 59

ACM Sigarch

MARCH 25, 2019

Part 1 previously discussed Bottleneck Analysis and Little’s Law, while this post (Part 2) presents the M/M/1 Queue. It also presented Bottleneck Analysis and Little’s Law that can give initial answers to questions like: What is the maximum throughput through several subsystems in series and parallel? and 1/S as ?.

Systems 53

Systems Latency Performance C++ 53

InfoQ

APRIL 15, 2024

The presenter shared the lessons learned from evolving and operating the platform, including cluster management and library versioning. At the QCon London 2024 conference, Félix GV from LinkedIn discussed the AI/ML platform powering the company’s products. By Rafal Gancarz

Artificial Intelligence 106

Artificial Intelligence Big Data Data Engineering Latency 106

All Things Distributed

JANUARY 18, 2012

a Fast and Scalable NoSQL Database Service Designed for Internet Scale Applications. Today is a very exciting day as we release Amazon DynamoDB , a fast, highly reliable and cost-effective NoSQL database service designed for internet scale applications. Amazon DynamoDB offers low, predictable latencies at any scale. Comments ().

Scalability 167

Scalability Database Ecommerce Latency 167

Brendan Gregg

FEBRUARY 28, 2023

My personal opinion is that I don't see a widespread need for more capacity given horizontal scaling and servers that can already exceed 1 Tbyte of DRAM; bandwidth is also helpful, but I'd be concerned about the increased latency for adding a hop to more memory. Ford, et al., “TCP

Performance 52

Performance Latency Cache Virtualization 52

Alex Russell

MAY 9, 2022

This is a complex topic, but to borrow from a recent post , web performance expands access to information and services by reducing latency and variance across interactions in a session, with a particular focus on the tail of the distribution (P75+). Consistent performance matters just as much as low average latency.

Performance 86

Performance Latency Metrics Engineering 86

Alex Russell

MARCH 6, 2021

This 2GiB RAM, Android 9 stalwart features the all-too classic lines of a Quad-core A53 (1.4GHz, small mercies) CPU, tastefully presented in a charming 5.5" Chip design choices and silicon economics are the defining feature of the still-growing Performance Inequality Gap. So what did $150USD fetch in 2019? " package.

Performance 124

Performance Network Cache Metrics 124

The Morning Paper

JUNE 13, 2019

HotOS’19 is presenting me with something of a problem as there are so many interesting looking papers in the proceedings this year it’s going to be hard to cover them all! The following figure highlights how just one of these variables, batch size, impacts throughput and latency on ResNet50. HotOS’19. Usability expectations.

Hardware 58

Hardware Latency Serverless Energy 58

The Morning Paper

MAY 12, 2019

A typical architecture diagram for one of these services looks like this: Suitably armed with a set of benchmark microservices applications, the investigation can begin! The top line shows the change in tail latency across a set of monolithic applications as operating frequency decreases. Hardware implications.

Open Source 61

Open Source Hardware Benchmarking Systems 61

Smashing Magazine

AUGUST 3, 2023

These pages serve as a pivotal tool in our digital marketing strategy, not only providing valuable information about our services but also designed to be easily discoverable through search engines. Large preview ) The graph presents the LCP values for the 75th percentile of our users visiting the Route Landing Page group.

Performance 74

Performance Cache Traffic Metrics 74

Percona

MAY 15, 2023

In short, any data that is written must be consistent with the data present. MySQL, as well as other RDBMS, are designed to work respecting the model and cannot scale in any way by fragmenting and distributing a schema, so what can be done to scale? This action is something that can affect very large datasets and be very expensive.

Traffic 121

Traffic Scalability Database Servers 121

cdemi

JUNE 10, 2017

Simply put, a Byzantine Fault is a fault that presents different symptoms to different observers. Several system architectures were designed that implement Byzantine Fault Tolerance. Because these are real-time systems, their Byzantine fault tolerance solutions must have very low latency.

Blockchain 62

Blockchain Latency Systems C++ 62

The Netflix TechBlog

DECEMBER 17, 2019

Designed with High Availability in mind. We designed the framework to minimize database impact. Go to step 1 if more chunks present. DBLog High Level Architecture. High Availability DBLog uses active-passive architecture. Supports any kind of output , so that the output can be a stream, datastore, or even an API.

Database 197

Database Traffic Transportation Open Source 197