Brendan Gregg

DECEMBER 30, 2017

My last talk for 2017 was at AWS re:Invent, on "How Netflix Tunes EC2 Instances for Performance," an updated version of my [2014] talk. WARNING: These tunables were developed in late 2017, for Ubuntu Xenial instances on EC2.** Networking. We help where we can. My talk included a section on Linux kernel tunables, as follows.

Tuning 52

Tuning AWS Best Practices Network 52

Brendan Gregg

SEPTEMBER 5, 2017

At my employer we sometimes use SR-IOV for direct network interface access, and NVMe for direct disk access. Most of the changes since rc7 are actually networking fixes, the bulk of them to various drivers. Live kernel patching is another capability, that is currently being integrated in the 4.x

Virtualization 75

Virtualization AWS Engineering Hardware 75

Dynatrace

MAY 6, 2021

Linux: Amazon Linux AMI 2017.x. On the “Network overview” page, `Host group` filter suggestions are now properly displayed when a management zone is selected. Operating systems support. Future Dynatrace Managed operating systems support changes. x86 (64bit-only). Vendor announcement. Linux: openSUSE 12.x, APM-292756).

Operating System 238

Operating System AWS Metrics Storage 238

Alex Russell

MARCH 6, 2021

TL;DR: A lot has changed since 2017 when we last estimated a global baseline resource per-page resource budget of 130-170KiB. To update our global baseline from 2017, we want to update our priors on a few dimensions: The evolved device landscape. Modern network performance and availability. The Moto G4 , for example. Hard Reset.

Performance 127

Performance Network Metrics Cache 127

Alex Russell

DECEMBER 18, 2022

TL;DR : To serve users at the 75 th percentile ( P75 ) of devices and networks, we can now afford ~150KiB of HTML/CSS/fonts and ~300-350KiB of JavaScript (gzipped). This is a slight improvement on last year's budgets , thanks to device and network improvements. Networks #. This is an ethical crisis for the frontend.

Performance 120

Performance Mobile Network Latency 120

Scalegrid

JULY 13, 2020

Recognized as the fastest growing database by popularity, PostgreSQL was named the DBMS of the year in both 2018 and 2017 by DB-Engines, and continues to grow in popularity in 2019. PostgreSQL is an open source object-relational database system with over 30 years of active development. In fact, PostgreSQL is so popular, 11.5%

Open Source 244

Open Source Tuning C++ Database 244

The Netflix TechBlog

JANUARY 22, 2021

channel surround sound in 2010, Dolby Atmos in 2017 , and adaptive bitrate audio in 2019. Using an approach similar to that described in our High Quality Audio Article , our xHE-AAC streams deliver studio-quality audio when network conditions allow, and minimize rebuffers when the network is congested. We began streaming 5.1

Programming 276

Programming Mobile Testing Network 276

O'Reilly

FEBRUARY 18, 2020

Up until 2017, the ML+AI topic had been amongst the fastest growing topics on the platform. After several years of steady climbing—and after outstripping Java in 2017—Python-related interactions now comprise almost 10% of all usage. Not necessarily: Java-related searches increased by 5% between 2017 and 2018. Coincidence?

Software Architecture 136

Software Architecture DevOps Data Engineering Architecture 136

All Things Distributed

NOVEMBER 13, 2018

Since then, AWS has added another PoP in Palermo in 2017. This helps support over 50 million searches every month on its networks across Italy. AWS also has a vibrant partner ecosystem in Italy as part of the AWS Partner Network (APN).

AWS 165

AWS Energy Automotive Traffic 165

Dynatrace

JUNE 15, 2023

If your application runs on servers you manage, either on-premises or on a private cloud, you’re responsible for securing the application as well as the operating system, network infrastructure, and physical hardware. What are some key characteristics of securing cloud applications? These alerts tend to slow down development.

Cloud 189

Cloud Open Source Architecture Monitoring 189

The Netflix TechBlog

OCTOBER 14, 2024

Blame The Network The next theory was that the network between the web browser UI (on the laptop) and the JupyterLab server was slow. the JupyterLab process) rather than the network. The UI apparently calls this function periodically , according to the network traffic tab in Chrome’s Developer Tools.

Latency 217

Latency Virtualization Traffic Processing 217

ACM Sigarch

APRIL 26, 2018

They unveiled their first two embedded AI chips fabricated with TSMC 40nm process in December 2017: “Journey 1.0 They use the graph as the basic representation for many AI-related algorithms, including neural network, Bayesian network, Markov Field, and some other emerging methods. CPU: Central Processing Unit.

Hardware 99

Hardware Artificial Intelligence Architecture Google 99

All Things Distributed

OCTOBER 25, 2018

That's where we built many pioneering networking technologies, our next-generation software for customer support, and the technology behind our compute service, Amazon EC2. In 2017, we brought the Amazon Global Network to Africa, through AWS Direct Connect. Since then, we have seen an acceleration in AWS adoption.

AWS 167

AWS Cloud Education Programming 167

Brendan Gregg

NOVEMBER 28, 2017

In this configuration, the AMI and boot is paravirt (PV), the kernel is making hypercalls instead of privileged instructions, and the system is using paravirt network and storage drivers. But not all workloads: some are network bound (proxies) and storage bound (databases). ## 5. AWS called this [enhanced networking].

Virtualization 40

Virtualization AWS Hardware Storage 40

The Netflix TechBlog

SEPTEMBER 13, 2021

Written by Jose Fernandez , Arthur Gonigberg , Julia Knecht , and Patrick Thomas In 2017, Netflix Studios was hitting an inflection point from a period of merely rapid growth to the sort of explosive growth that throws “how do we scale?” into every conversation.

Internet 229

Internet Internet Cloud Traffic 229

All Things Distributed

MAY 19, 2017

We believe the central constraint in high throughput data processing has moved from compute and storage to the network. We describe how doing so not only reduces network traffic, but also allows for fast crash recovery, failovers to replicas without loss of data, and fault-tolerant, self-healing storage.

Database 91

Database Design Cloud Storage 91

The Netflix TechBlog

JULY 1, 2019

An example console output showing surface cache over budget errors During brainstorming with our platform team, we came across an internal bug report from 2017 that described a similar issue to what we were seeing?—?surfaces those running the 2017 SDK?—?also The SNG files have a relatively large download size at ~1.5MB each.

Cache 167

Cache Design Testing Development 167

Brendan Gregg

FEBRUARY 28, 2023

Titus, the Netflix container management platform, is now open source,” [link] Apr 2018 - [Cutress 19] Dr. [link] 2021 - [Trader 21] Tiffany Trader, “Cerebras Second-Gen 7nm Wafer Scale Engine Doubles AI Performance Over First-Gen Chip ,” [link] Apr 2021 - [Ghigoff 21] Yoann Ghigoff, et al., "BMC:

Performance 103

Performance Latency Cache Virtualization 103

All Things Distributed

SEPTEMBER 25, 2017

I'm also excited to announce today that we are launching an AWS Edge Network Location in the United Arab Emirates (UAE) in the first quarter of 2018. 2017 continues a busy year for AWS in the Middle East. We still have another five AWS Regions (and 14 Availability Zones) in China, France, Hong Kong, and Sweden.

Cloud 122

Cloud Education Energy Government 122

Dynatrace

OCTOBER 17, 2019

Everything as Code can be described as a methodology or practice which consists of extending the idea of how applications are treated as code and applying these concepts to all other IT components like operating systems, network configurations, and pipelines. We went from managing five EC2 instances in 2011 to around 1,000 in 2017.

Code 112

Code Infrastructure DevOps Cloud 112

Alex Russell

JANUARY 30, 2024

It's time once again to update our priors regarding the global device and network situation. seconds on the target device and network profile, consuming 120KiB of critical path resources to become interactive, only 8KiB of which is script. What's changed since last year? and 75KiB of JavaScript. These are generous targets.

Performance 98

Performance Network Mobile Speed 98

The Symphonia

DECEMBER 6, 2017

to a massively distributed CDN network a possibility. Serverless at re:Invent 2017 was originally published in The Symphonium on Medium, where people are continuing the conversation by highlighting and responding to this story. We continue to be fascinated to see how people will use Lambda@Edge over the coming years.

Serverless 40

Serverless Lambda AWS Traffic 40

Tim Kadlec

JANUARY 9, 2019

Sites that use an excess of resources, whether on the network or on the device, don’t just cause slow experiences, but can leave entire groups of people out. I recently profiled a page on a Pixel 2 (released in 2017), and an Alcatel 1x (released in 2018). Poor performance can, and does, lead to exclusion.

Performance 111

Performance Energy Internet Internet 111

All Things Distributed

DECEMBER 12, 2018

In April 2017, Amazon Web Services announced that it would launch a new AWS infrastructure region Region in Sweden. We launched Edge Network locations in Denmark, Finland, Norway, and Sweden. The AWS Europe (Stockholm) Region is our fifth European Region, joining Dublin, Frankfurt, London, and Paris.

AWS 120

AWS Cloud Games Serverless 120

Tim Kadlec

JANUARY 4, 2018

It’s not much of a surprise, then, that I read more books in 2017 than I have since I started keeping track back in 2009. His point, very Neil Postman-esque, is particularly true in today’s world of entertainment-driven news and fast-moving social networks that I’m increasingly convinced that, by default, only amplify existing biases.

BioTech 45

BioTech Social Media Entertainment Media 45

Scalegrid

AUGUST 27, 2024

Scalability Strategies for IoT Networks As the number of devices connected to IoT networks expands, ensuring scalability becomes crucial. What scalability features does RabbitMQ offer for IoT networks? Consistent hashing and horizontal scaling are robust scalability features RabbitMQ provides for IoT networks.

IoT 130

IoT Ecommerce Games Scalability 130

SQL Server According to Bob

MAY 19, 2019

The reduction effort started when we shipped SQL Server 2017 ( [link].) Networking Pump Threads. SQLPAL uses a dynamic set of networking, I/O pump threads to handle network requests. SQLPAL uses a dynamic set of async, I/O pump threads to handle non-network I/O requests. SQL Server 2017. Async Pump Threads.

IoT 61

IoT Servers Development Cache 61

Tim Kadlec

JUNE 21, 2022

HTTP/3 work started in 2012 with Google working on QUIC, adopted by IETF in 2017, RFC’s published in June 2022. The TLS + QUIC layer makes UDP safer to use, but a lot of networks will likely still block for awhile. The RSVP problem: HTTP/3 might be blocked on the network or may not be enabled on the server. What and Why?

Network 65

Network Mobile Google Cache 65

Tasktop

APRIL 2, 2020

from Project to Product (based on Perez 2017 ) . from Project to Product (based on Perez 2017 ) showing the technological revolutions, with added overlay to show the events that initiated each of the Turning Points . Perez’s models break each of the five technological revolutions into three distinct phases.

Government 109

Government Innovation Strategy Technology 109

All Things Distributed

OCTOBER 2, 2017

In support of Amazon Prime Day 2017, the biggest day in Amazon retail history, DynamoDB served over 12.9 VPC Endpoints give you the ability to control whether network traffic between your application and DynamoDB traverses the public Internet or stays within your virtual private cloud. million requests per second.

Internet 113

Internet Internet AWS Performance 113

Brendan Gregg

JULY 15, 2019

At over 700 pages, it covers CPUs, memory, disks, file systems, networking, multiple languages, applications, containers, hypervisors, security analysis, and the Linux kernel. - It also covers over 30 traditional tools (iostat(1), perf(1), etc) so you can use the right tool for the job, even when that tool is not BPF. -

Systems 111

Systems Performance Education Technology 111

Smashing Magazine

AUGUST 10, 2023

The other issue is that the data I’m getting back is based on lab simulations where I can add throttling, determine the device that’s used, and the network connection, among other simulated conditions. On that note, it’s worth calling out that there are multiple flavors of network throttling. Real usage data would be better, of course.

Speed 82

Speed Monitoring Testing Network 82

MachMetrics

FEBRUARY 7, 2019

Facebook took notice of the 3-second abandonment statistic and declared in August 2017 that information feed updates with slow-loading links will be shown reduced in the newsfeed: “With this update, we’ll soon take into account the estimated load time of a webpage that someone clicks to from any link in News Feed on the mobile app.

Speed 93

Speed Website Performance Website Mobile 93

Brendan Gregg

JULY 4, 2021

AWS Graviton2); for memory with the arrival of DDR5 and High Bandwidth Memory (HBM) on-processor; for storage including new uses for 3D Xpoint as a 3D NAND accelerator; for networking with the rise of QUIC and eXpress Data Path (XDP); and so on.

Performance 52

Performance Latency Hardware Storage 52

Dotcom-Montior

JUNE 25, 2020

A Content Delivery Network (CDN) is a network of servers that is placed in different locations across the globe. Between June 2017 and 2018, the total number of SSL certificates increased by a record 13.2 Subject key identifier : This helps the certificate to be recognized as containing that particular public key.

Monitoring 77

Monitoring Website Internet Internet 77

Alex Russell

OCTOBER 22, 2017

We constrain ourselves to a real-world baseline device + network configuration to measure progress. — Alex Russell (@slightlylate) October 4, 2017. Budgets are scaled to a benchmark network & device. For developers and decision makers with fast phones on fast networks this is a double-whammy of hidden costs.

Performance 74

Performance Network Benchmarking Mobile 74

CSS - Tricks

APRIL 1, 2021

A 2017 study by Akamai says as much when it found that even a 100ms delay in page load can decrease conversions by 7% and lose 1% of their sales for every 100ms it takes for their site to load which, at the time of the study, was equivalent to $1.6 In DevTools, the Network inspector helps us see what the first webpage is doing too.

Cache 92

Cache Social Media Media Network 92