Brendan Gregg

FEBRUARY 28, 2023

This talk originated from my updates to [Systems Performance 2nd Edition], and this was the first time I've given this talk in person! CXL in a way allows a custom memory controller to be added to a system, to increase memory capacity, bandwidth, and overall performance. Ford, et al., “TCP

Performance 103

Performance Latency Cache Virtualization 103

All Things Distributed

APRIL 4, 2017

The new AWS EU (Stockholm) Region will have three Availability Zones and will be ready for customers to use in 2018. After finding it cost prohibitive to use colocation centers in local markets where their users are based, iZettle decided to give up hardware. Over the past decade, we have seen tremendous growth at AWS.

AWS 107

AWS Airlines Latency Games 107

O'Reilly

JANUARY 13, 2020

” Willing also offered a shout-out to the CircuitPython and Mu projects, asking, “Who doesn’t love hardware, blinking LEDs, sensors, and using Mu, a user-friendly editor that is fantastic for adults and kids?” ” Java. It’s mostly good news on the Java front.

Programming 144

Programming Java Google C++ 144

All Things Distributed

NOVEMBER 21, 2018

Clearly, they're updating their systems far more frequently than they did on-premises, which changes the nature of engineering problems the team needs to prioritize. Another significant difference between on-premises systems and the cloud is the abundance of available resources. Optimizing performance for bursts of user activity.

Cloud 115

Cloud Benchmarking Performance AWS 115

The Morning Paper

MARCH 1, 2020

Firecracker is the virtual machine monitor (VMM) that powers AWS Lambda and AWS Fargate, and has been used in production at AWS since 2018. Firecracker (and Cloud Hypervisor) is limited to around 13,000 IOPS (52 MB/s at 4kB), whereas the same hardware is capable of over 340,000 read IOPS (1GB/s at 4kB).

Virtualization 101

Virtualization Serverless Lambda AWS 101

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. Pointer Lock. Media Recorder.

Media 145

Media Games Education Engineering 145

All Things Distributed

DECEMBER 12, 2018

In 2018, we launched the first Pop-Up Loft in Stockholm, offering a co-working space and access to technology and business experts to support the growth of Nordic organizations. We help Supercell to quickly develop, deploy, and scale their games to cope with varying numbers of gamers accessing the system throughout the course of the day.

AWS 119

AWS Cloud Games Serverless 119

O'Reilly

NOVEMBER 28, 2023

That pricing won’t be sustainable, particularly as hardware shortages drive up the cost of building infrastructure. These models are typically smaller (7 to 14 billion parameters) and easier to fine-tune, and they can run on very limited hardware; many can run on laptops, cell phones, or nanocomputers such as the Raspberry Pi.

Open Source 135

Open Source Google Artificial Intelligence Infrastructure 135

Randon ASCII

FEBRUARY 25, 2018

See the end of the post for an October 2018 bug fix update, or read the whole story: Flaky failures are the worst. In this particular investigation, which spanned twenty months, we suspected hardware failure, compiler bugs, linker bugs, and other possibilities. Update, October 2018.

Programming 79

Programming Hardware Cache Code 79

Brendan Gregg

JULY 4, 2021

This was a chance to talk about other things I've been working on, such as the present and future of hardware performance. I also wrote about these topics in detail for my recent [Systems Performance 2nd Edition] book. Note that my predictions in this talk may be wrong, but they should be thought-provoking. Ford, et al., “TCP

Performance 52

Performance Latency Hardware Storage 52

Nick Desaulniers

OCTOBER 24, 2018

Before booting a compiled Linux kernel image on actual hardware, it can save us time and potential headache to do a quick boot in a virtual machine like QEMU. qemu-system-x86_64. qemu-system-x86_64 -kernel arch/x86_64/boot/bzImage. qemu-system-x86_64 -kernel arch/x86_64/boot/bzImage -nographic. as a sanity check.

C++ 56

C++ Hardware Virtualization Systems 56

Alex Russell

FEBRUARY 21, 2023

Recall that Cupertino manages the actual work of Safari engineers through Apple-internal systems (previously named "Radar" ), making public bug reports a sort of parallel track; once an issue is imported from a public bug to a private tracker, it's more likely to get developer attention.

Energy 90

Energy Games Engineering Design 90

Abhishek Tiwari

DECEMBER 27, 2017

In 2018, a widespread adaptation of Kubernetes for big data processing is anitcipated. A good example will be something like Pachyderm [8] [9] - a relatively young project which implements a distributed Pachyderm File System (PFS) and a data-aware scheduler Pachyderm Pipeline System (PPS) on top of Kubernetes.

Big Data 63

Big Data Storage Benchmarking Hardware 63

HammerDB

OCTOBER 12, 2018

As with the previous guides as an Intel employee (#IAMINTEL) the examples are taken from a MySQL 8 on Linux on Intel system and the approach is the same for whatever system you are testing although some of the settings you see may be different. Similarly for this guide MySQL can be swapped for a mySQL based databases such as MariaDB.

Best Practices 40

Best Practices Scalability Performance C++ 40

Percona

AUGUST 29, 2018

For optimum performance, a PostgreSQL database depends on the operating system parameters being defined correctly. PostgreSQL uses System V (SysV) that requires SHMMAX setting. So now it requires fewer bytes of System V shared memory. DETAIL: Failed system call was shmget(key=1, size=2072576, 03600). Until version 9.2,

Tuning 79

Tuning Database Servers Processing 79

Brendan Gregg

AUGUST 18, 2019

BPF has been extended (aka eBPF) in recent years for providing a safe way to extend kernel functionality, and has become a hot topic in systems engineering, with at least 24 talks on BPF at the last Linux Plumber's conference. hardware Hardware counter-based instrumentation. system(".") BEGIN Start of bpftrace.

Latency 57

Latency C++ Cache Programming 57

Brendan Gregg

FEBRUARY 28, 2023

This talk originated from my updates to Systems Performance 2nd Edition , and this was the first time I've given this talk in person! CXL in a way allows a custom memory controller to be added to a system, to increase memory capacity, bandwidth, and overall performance. Ford, et al., “TCP

Performance 52

Performance Latency Cache Virtualization 52

HammerDB

OCTOBER 8, 2018

This post gives a HOWTO guide on system configuration for achieving top levels of performance with the HammerDB PostgreSQL TPC-C test. Firstly for system choice a 2 socket system is optimal for PostgreSQL OLTP performance at the time of writing. CPUs which run at the same hardware frequency: 0. CPU Configuration.

Best Practices 40

Best Practices Scalability Performance Hardware 40

KeyCDN

SEPTEMBER 20, 2018

Let’s take a look at some of the best error monitoring tools and strategies for 2018. Monitoring errors on the front-end requires a bit more work because front-end performance is highly dependant on the user’s hardware, software and connection. What is Error Tracking?

Monitoring 67

Monitoring Strategy Website Google 67

Smashing Magazine

FEBRUARY 5, 2019

Although modern iOS hardware is powerful enough to handle many intensive and complex tasks, the device could still feel unresponsive if you are not careful about how your app performs. The app snapshots we see in the multitasking view are actually screenshots taken by the system from right when we exited the app (i.e.

Performance 89

Performance Energy Operating System Code 89

Percona

MAY 1, 2023

This blog was originally published in August 2018 and was updated in May 2023. It is primarily the responsibility of the database administrator or developer to tune PostgreSQL according to their system’s workload. Out of the box, the default PostgreSQL configuration is not tuned for any particular workload.

Tuning 52

Tuning Database Efficiency Performance 52

Alex Russell

OCTOBER 22, 2017

Tools and CI systems help them monitor progress & prevent regressions. Some systems show median RTTs near ~100ms for 3G users. Given the conflicted data we see across our other systems, this seems about right as a baseline. Think back: your framework of choice might be 40K, but what about that data system?

Performance 74

Performance Network Benchmarking Mobile 74

Alex Russell

JANUARY 30, 2024

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. Since 2010, volumes have been on a slow downward glide path , shrinking from ~350MM per year in a decade ago to ~260MM in 2018. was a system properly localised?

Performance 98

Performance Network Mobile Speed 98

Randon ASCII

AUGUST 16, 2018

As usual I had UIforETW running in the background, tracing to circular buffers, so I just had to type Ctrl+Win+R and the buffers, representing the last thirty seconds or so of system activity, were saved to disk. I loaded it into Windows Performance Analyzer (WPA), but I couldn’t definitively find the hang. But it’s not. Source code.

Programming 75

Programming Virtualization Code Processing 75

Percona

JULY 1, 2023

This post was originally published in July 2018 and was updated in July 2023. Understanding DBaaS DBaaS cloud services allow users to use databases without configuring physical hardware and infrastructure or installing software. What are the differences between Aurora and RDS? How do I choose which one to use?

AWS 52

AWS Database Serverless Storage 52

High Scalability

SEPTEMBER 14, 2018

editingemily : "In peacetime, a ticketing system is great. Because of that big servers and other memory systems need to have another kind of memory in the hierarchy. Thus, despite both quoting 11 nines of durability against hardware failures, S3 is durable against failures that B2 is not, and is thus better. @RachFrieee : 3.6B

Internet 30

Internet Internet Scalability Education 30

ACM Sigarch

APRIL 26, 2018

The new DGX-2, which achieves 2 petaFLOPS in the system, combines 16 fully interconnected GPUS with 10x the deep learning performance. HPU: Holographic Processing Unit (HPU) is the specific hardware of Microsoft’s Hololens. More details can be found in their recent talk in Scaled Machine Learning in March 2018.

Hardware 100

Hardware Artificial Intelligence Architecture Google 100

High Scalability

JANUARY 4, 2019

It's HighScalability time: Solar system? By replacing branch-heavy algorithms with neural networks, the DBMS can profit from these hardware trends.". Nope, the beauty is in your head— neural art. Do you like this sort of Stuff? Please support me on Patreon. Need cloud? Explain the Cloud Like I'm 10 (34 almost 5 star reviews).

Internet 29

Internet Internet Scalability Healthcare 29

SQL Server According to Bob

DECEMBER 18, 2018

Durability: “In database systems , durability is the ACID property which guarantees transactions that have committed will survive permanently. For example, if a flight booking reports that a seat has successfully been booked, then the seat will remain booked even if the system crashes.” – [link]. The Back Story.

Servers 90

Servers Media Cache Storage 90

CSS Wizardry

APRIL 2, 2018

— Harry Roberts (@csswizardry) 18 March, 2018. pic.twitter.com/2Pm6Gt7NXq — Harry Roberts (@csswizardry) March 20, 2018. — Harry Roberts (@csswizardry) March 21, 2018. Know anywhere I can get hold of a ‘dubious’ copy of this track? I’ll donate £25 to a charity of your choosing (DMs are open). Combine Libraries.

Website 109

Website Media Hardware Monitoring 109

Alex Russell

MARCH 6, 2021

India became a 4G-centric market sometime in 2018. Hardware Past As Performance Prologue. Regardless, the overall story for hardware progress remains grim, particularly when we recall how long device replacement cycles are: Tap for a larger version. 5G looks set to continue a bumpy rollout for the next half-decade. Mind The Gap.

Performance 127

Performance Network Mobile Metrics 127

SQL Performance

JULY 16, 2019

A close monitoring of the hardware enthusiast community, including many of the most respected hardware analysts and reviewers paints an even more dire picture about Intel in the server processor space. This made it easier for database professionals to make the case for a hardware upgrade, and made the typical upgrade more worthwhile.

Servers 46

Servers Hardware Architecture Retail 46

John McCalpin

JANUARY 22, 2018

Dr. Damon McDougall gave a short presentation on this study at the IXPUG 2018 Fall Conference ( pdf ) — I originally wrote these notes to help organize my thoughts as we were preparing the IXPUG presentation, and later decided that the extra details contained here are interesting enough for me to post it.

Latency 40

Latency Hardware Code Testing 40

John McCalpin

JANUARY 22, 2018

Dr. Damon McDougall gave a short presentation on this study at the IXPUG 2018 Fall Conference ( pdf ) — I originally wrote these notes to help organize my thoughts as we were preparing the IXPUG presentation, and later decided that the extra details contained here are interesting enough for me to post it. cmpl $1000000000, %eax.

Latency 40

Latency Hardware Code Testing 40

Smashing Magazine

JANUARY 15, 2019

The Mozilla Internet Health Report 2018 states that — especially as the Internet expands into new territory — “sustainability should be a bigger priority.” how much data does the browser have to download to display your website) and resource usage of the hardware serving and receiving the website.

Website Performance 78

Website Performance Website Energy Performance 78

Smashing Magazine

JANUARY 21, 2021

We never used CSS-in-JS or styled-components, but instead a good ol’ component-based system of Sass-modules which would be compiled into CSS. The reasons for it are numerous, but the most important one is a huge difference in network conditions and device hardware across the world. KB, Brotlified). So what’s then?

Performance 111

Performance Cache Mobile Network 111

Smashing Magazine

JANUARY 7, 2019

On the other hand, we have hardware constraints on memory and CPU due to JavaScript parsing times (we’ll talk about them in detail later). Gatsby.js (React), Preact CLI , and PWA Starter Kit provide reasonable defaults for fast loading out of the box on average mobile hardware. But this consideration needs to be deliberate.

Performance 111

Performance Cache Network Metrics 111