The Morning Paper

JUNE 27, 2019

Machine learning systems are stuck in a rut Barham & Isard, HotOS’19. In this paper we argue that systems for numerical computing are stuck in a local basin of performance and programmability. Take a simple example: it would be really nice if we could have named dimensions instead of always having to work with indices.

Systems 87

Systems Tuning Programming Innovation 87

Percona

JUNE 22, 2023

Defining high availability In general terms, high availability refers to the continuous operation of a system with little to no interruption to end users in the event of hardware or software failures, power outages, or other disruptions. Some disruption might occur, but it will be minimal. More in the following sub-section.)

Availability 107

Availability Database Open Source Hardware 107

Dynatrace

SEPTEMBER 30, 2021

It requires purchasing, powering, and configuring physical hardware, training and retaining the staff capable of servicing and securing the machines, operating a data center, and so on. They need enough hardware to serve their anticipated volume and keep things running smoothly without buying too much or too little.

Cloud 162

Cloud Traffic Best Practices Strategy 162

Testsigma

MAY 3, 2021

From operating systems to versions to hardware specs, mobile devices stand unique even after being billions in number. But testing this is essential too because data communication over the network cannot be eliminated by putting all the files into the device storage. Training and Selecting Right Tool Stack.

Mobile 77

Mobile Testing Operating System Hardware 77

Adrian Cockcroft

JUNE 2, 2023

One example is that Netflix launched it’s streaming service in 2007, just at the point when the cost of streaming a movie over the network (which was dropping fast) became less than the cost of shipping a DVD. You need to be able to innovate fast enough to pivot or reinvent your business model and leverage the change.

Social Media 98

Social Media Innovation Open Source Entertainment 98

Percona

OCTOBER 18, 2023

Parallel run : Operating both the existing and new database system simultaneously during the transition, allowing for a seamless fallback in case of any issues. Resource allocation: Personnel, hardware, time, and money The migration to open source requires careful allocation (and knowledge) of the resources available to you.

Open Source 91

Open Source Database Hardware Strategy 91

The Morning Paper

MAY 14, 2019

Last time around we looked at the DeathStarBench suite of microservices-based benchmark applications and learned that microservices systems can be especially latency sensitive, and that hotspots can propagate through a microservices architecture in interesting ways. Distributed tracing and instrumentation.

Big Data 45

Big Data Cloud Performance Hardware 45

The Morning Paper

JUNE 13, 2019

As we saw with the SOAP paper last time out, even with a fixed model variant and hardware there are a lot of different ways to map a training workload over the available hardware. Managed here means that the system automates resource provisioning for models to match a set of SLO constraints (cf. autoscaling).

Hardware 58

Hardware Latency Serverless Energy 58

The Morning Paper

FEBRUARY 27, 2020

Modern software systems at scale are incredibly complex ever changing environments. memory leaks that take hours to build up into an issue); and there can be problems that only exhibit themselves with certain user, hardware, or software configurations. Gandalf system design. NSDI’20.

Analytics 68

Analytics Infrastructure Cloud Hardware 68

The Morning Paper

MAY 28, 2019

We don’t talk about energy as often as we probably should on this blog, but it’s certainly true that our data centres and various IT systems consume an awful lot of it. The goal is to produce a low-energy hardware classifier for embedded applications doing local processing of sensor data. Introducing race logic.

Energy 52

Energy Hardware Efficiency Architecture 52

Adrian Cockcroft

JUNE 21, 2022

Photo by Adrian I spent six years at Cambridge Consultants, building some interesting systems, managing our Sun workstations and learning a lot, but by then Sun had opened a sales office across the street, and I wanted to find out what they were going to release next, before everyone else.

Tuning 52

Tuning Benchmarking Engineering C++ 52

Alex Russell

JANUARY 30, 2024

It's time once again to update our priors regarding the global device and network situation. HTML, CSS, images, and fonts can all be parsed and run at near wire speeds on low-end hardware, but JavaScript is at least three times more expensive, byte-for-byte. What's changed since last year? and 75KiB of JavaScript.

Performance 90

Performance Network Mobile Speed 90

Sutter's Mill

FEBRUARY 13, 2017

” This contains updated and new material that reflects the latest C++ standards and compilers, with a focus to using modern C++11/14/17 effectively on modern hardware and memory architectures. On April 25-27, I’ll be in Stockholm (Kista) giving a three-day seminar on “High-Performance and Low-Latency C++.”

Latency 51

Latency C++ Hardware Performance 51

The Morning Paper

FEBRUARY 20, 2020

Raven is the system that Microsoft built to explore this question, and answer it with a resounding yes. NN translation optimisations replace classical ML operators and data ‘featurizers’ with neural networks that can be executed directly in e.g. ONNX Runtime, PyTorch, or TensorFlow. " " Query execution.

Processing 72

Processing Hardware Database Azure 72

ACM Sigarch

APRIL 8, 2019

However, we often hear anecdotes that the number of prospective graduate students applying to computer architecture/systems is small and shrinking. This is arguably a fundamentally hard problem for computer architecture, but efforts towards open source hardware (eg. Why is that? Should we be alarmed as a community? Lack of Diversity.

Architecture 74

Architecture Open Source Hardware Software Engineering 74

The Morning Paper

JULY 18, 2019

Being an ICLR paper of course, you won’t be surprised to learn that the solution involves neural networks. Instead of learning from scratch, we draw inspiration from the few-shot leaning advances obtained by meta-learning memory-augmented neural networks.

Network 86

Network Java Database Hardware 86

Alex Russell

APRIL 29, 2021

After 20 years of neck-in-neck competition, often starting from common code lineages, there just isn't that much left to wring out of the system. Consistent improvement is the name of the game, and it can still have positive impacts, particularly as users lean on the system more heavily over time. Delayed five years. Gamepad API.

Media 145

Media Games Education Engineering 145

Alex Russell

JUNE 20, 2020

Recall that browsers arose atop platforms that universally provided a host of services by the late 90's: system fonts (to support layout programs, including WYSIWYG editing). full TCP/IP & UDP networking, including local network discovery. In each quantum of time, hardware and OS vendors press ahead, adding features.

Hardware 55

Hardware Games Java Media 55

O'Reilly

NOVEMBER 28, 2023

Training models and developing complex applications on top of those models is becoming easier. Many of the new open source models are much smaller and not as resource intensive but still deliver good results (especially when trained for a specific application). report that the difficulty of training a model is a problem.

Open Source 108

Open Source Artificial Intelligence Google Infrastructure 108

O'Reilly

SEPTEMBER 19, 2023

A company then needed to train up their ops team to manage the cluster, and their analysts to express their ideas in MapReduce. Doubly so as hardware improved, eating away at the lower end of Hadoop-worthy work. And that brings our story to the present day: Stage 3: Neural networks High-end video games required high-end video cards.

Hardware 99

Hardware Storage Big Data Blockchain 99

High Scalability

OCTOBER 24, 2022

There is no structural separation as there is with a train service. Each cloud-native evolution is about using the hardware more efficiently. Network effects are not the same as monopoly control. Anyone can rent a rack in a colo and stand up a system. And even that list is not invulnerable. Neither are clouds.

Cloud 187

Cloud Hardware Infrastructure Innovation 187

The Netflix TechBlog

APRIL 29, 2019

Open Connect Open Connect is Netflix’s content delivery network (CDN). video streaming) takes place in the Open Connect network. Various software systems are needed to design, build, and operate this CDN infrastructure, and a significant number of them are written in Python. are you logged in?

Open Source 279

Open Source Network Infrastructure Big Data 279

O'Reilly

AUGUST 4, 2022

DeepMind’s Gato is an AI model that can be taught to carry out many different kinds of tasks based on a single transformer neural network. The achievement of note is that it’s underpinned by a single model trained across all tasks rather than different models for different tasks and modalities.

Games 104

Games Social Media Hardware Architecture 104

The Netflix TechBlog

DECEMBER 23, 2020

By Fabio Kung , Sargun Dhillon , Andrew Spyker , Kyle , Rob Gulewich, Nabil Schear , Andrew Leung , Daniel Muino, and Manas Alekar As previously discussed on the Netflix Tech Blog, Titus is the Netflix container orchestration system. everything from the frontend API for netflix.com, to machine learning training workloads, to video encoders.

Media 286

Media Metrics Systems Processing 286

All Things Distributed

MARCH 19, 2018

For example, training on more data means more accurate models. Last re:Invent, to make the problem of authoring, training, and hosting ML models easier, faster, and more reliable, we launched Amazon SageMaker. Machine learning models are usually trained tens or hundreds of times. In machine learning, more is usually more.

Scalability 159

Scalability Hardware AWS Tuning 159

O'Reilly

OCTOBER 11, 2022

As a trend, it’s not performing well on Google; it shows little long-term growth, if any, and gets nowhere near as many searches as terms like “Observability” and “Generative Adversarial Networks.” meme originated in IT’s transformation from manual system administration to automated configuration management and software deployment.

Games 116

Games Healthcare Processing Retail 116

All Things Distributed

DECEMBER 12, 2018

We launched Edge Network locations in Denmark, Finland, Norway, and Sweden. We also provided web-based training, self-paced labs, customer support, third-party offers, and up to $100,000 in AWS service credits–all at no charge. Today, we add to that presence with an infrastructure Region in Stockholm with three Availability Zones.

AWS 124

AWS Cloud Games Serverless 124

Smashing Magazine

AUGUST 22, 2021

As we will see, QUIC and HTTP/3 indeed have great web performance potential, but mainly for users on slow networks. If your average visitor is on a fast cabled or cellular network, they probably won’t benefit from the new protocols all that much. An often used metaphor is that of a pipe used to transport water. Congestion Control.

Performance 106

Performance Network Latency Servers 106

O'Reilly

JULY 20, 2021

Most recently, IBM used Flickr photos that were shared under a Creative Commons license to train facial recognition algorithms. There may be alarm systems. The concept of Zero Trust Networks speaks to this problem. This is why people have made a business out of setting up home stereo and video systems.

Google 121

Google Games Technology Technology 121

The Agile Manager

JUNE 23, 2007

This might be a data centre where hardware uptime is guaranteed to process transactions, a timesheeting capability that is available on-demand, or development of a custom application to analyse asset backed securities. Traditionally, IT has been a risk assumer. staff turnover, knowledge continuity, individual skill, interpersonal factors.)

Technology 40

Technology Technology Energy Innovation 40

All Things Distributed

NOVEMBER 30, 2016

in ML and neural networks) and access to vast amounts of data. automatic speech recognition, natural language understanding, image classification), collect and clean the training data, and train and tune the machine learning models.

AWS 165

AWS Lambda Artificial Intelligence Mobile 165

High Scalability

NOVEMBER 25, 2019

He and his colleagues spend their productive hours scaling large distributed file systems. Our only option was to roll up our sleeves and build basic cloud file system components such as object store ourselves. Java used to power core file system code. What is the name of your system and where can we find out more about it?

Architecture 52

Architecture Cache Azure Storage 52