Brendan Gregg

MARCH 16, 2024

2015-2020: Overhead As part of production rollout I did many performance overhead tests, which I've described publicly before: The overhead of adding frame pointers to everything (libc and Java) was usually less than 1%, with one exception of 10%. The actual overhead depends on your workload. Others have reported around 1% and around 2%.

Java 145

Java Cache Google Hardware 145

Alex Podelko

SEPTEMBER 8, 2020

It was called Jellly at the time (mid 2015) but it was an early build of what OctoPerf would become. And when we started in 2015, JMeter was undoubtedly the best tool around. You have to remember this was in 2015, a lot has changed since, Gatling is definitely a stronger tool.

Open Source 223

Open Source Performance Testing Testing Performance 223

Alex Russell

FEBRUARY 21, 2023

In leaner years (2012-2015), a single Fall release was all we'd get. From outright misstatements about a competitor's security, to claims that performance differences in hardware show Safari is faster, to [geographic brinksmanship](/2022/02/minimum-standards/), the confident bluster hasn't gone down particularly well.

Energy 79

Energy Games Engineering Design 79

Brendan Gregg

NOVEMBER 29, 2017

Hardware virtualization for cloud computing has come a long way, improving performance using technologies such as VT-x, SR-IOV, VT-d, NVMe, and APICv. The latest AWS hypervisor, Nitro, uses everything to provide a new hardware-assisted hypervisor that is easy to use and has near bare-metal performance. I'd expect between 0.1%

Virtualization 111

Virtualization AWS Hardware Storage 111

Brendan Gregg

SEPTEMBER 25, 2021

As a Xen guest, this profile was gathered using perf(1) and the kernel's software cpu-clock soft interrupts, not the hardware NMI. I also shared setting the clocksource in my talks and in my 2015 [Linux tunables] post. Theory (A) is most likely based on the frame widths in the flame graph. But I'm not completely sure.

Speed 126

Speed Java AWS Virtualization 126

The Netflix TechBlog

DECEMBER 23, 2020

Unfortunately, these default namespace boundaries are not sufficient to prevent container escape, as seen in CVEs like CVE-2015–2925. In addition to the default Docker namespaces (mount, network, UTS, IPC, and PID), we employ user namespaces for added layers of isolation.

Media 286

Media Metrics Systems Processing 286

All Things Distributed

APRIL 4, 2017

In 2014 and 2015 respectively, AWS opened offices in Stockholm and Espoo, Finland. After finding it cost prohibitive to use colocation centers in local markets where their users are based, iZettle decided to give up hardware. In making the switch to AWS, WOW air has saved between $30,000 and $45,000 on hardware, and software licensing.

AWS 117

AWS Airlines Latency Games 117

Brendan Gregg

NOVEMBER 28, 2017

Hardware virtualization for cloud computing has come a long way, improving performance using technologies such as VT-x, SR-IOV, VT-d, NVMe, and APICv. The latest AWS hypervisor, Nitro, uses everything to provide a new hardware-assisted hypervisor that is easy to use and has near bare-metal performance. I'd expect between 0.1%

Virtualization 40

Virtualization AWS Hardware Storage 40

Brendan Gregg

SEPTEMBER 25, 2021

As a Xen guest, this profile was gathered using perf(1) and the kernel's software cpu-clock soft interrupts, not the hardware NMI. I also shared setting the clocksource in my talks and in my 2015 [Linux tunables] post. Theory (A) is most likely based on the frame widths in the flame graph. But I'm not completely sure.

Speed 40

Speed Java AWS Virtualization 40

Percona

JANUARY 17, 2019

On multi-core machines – which is the majority of the hardware nowadays – and in the cloud, we have multiple cores available for use. In all versions of MySQL – at least at the time of writing – when you run a single query it will run in one thread, effectively utilizing one CPU core only. With faster disks (i.e. sec). /* another run */.

Cache 46

Cache C++ AWS Airlines 46

SQL Server According to Bob

JUNE 3, 2016

My development collogues and I are starting a regular blog series, outlining the vast range of scalability improvements, allowing SQL Server 2016 to run across a wide array of hardware configurations, faster and better than previous releases of SQL Server. Try SQL Server 2016 Today.

Hardware 40

Hardware Software Engineering Transportation Scalability 40

Randon ASCII

FEBRUARY 25, 2018

In this particular investigation, which spanned twenty months, we suspected hardware failure, compiler bugs, linker bugs, and other possibilities. Jumping too quickly to blaming hardware or build tools is a classic mistake, but in this case the mistake was that we weren’t thinking big enough. Russian translation is here.

Programming 79

Programming Hardware Cache Code 79

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

Brendan Gregg

SEPTEMBER 25, 2021

As a Xen guest, this profile was gathered using perf(1) and the kernel's software cpu-clock soft interrupts, not the hardware NMI. I also shared setting the clocksource in my talks and in my 2015 Linux tunables post. Theory (A) is most likely based on the frame widths in the flame graph. But I'm not completely sure.

Speed 52

Speed Java AWS Virtualization 52

Randon ASCII

AUGUST 16, 2018

Aside: I just want to say thank you to 2015 me for asking Microsoft to improve their thread naming mechanisms, and thank you to Microsoft for doing everything I suggested – thread names in WPA are great ! The problem suddenly became clearer. Source code. Rabbit holes.

Programming 75

Programming Virtualization Code Processing 75

Alex Russell

MARCH 6, 2021

Devices and networks have evolved too: Alex Russell @slightlylate An update on mobile CPUs and the Performance Inequality Gap: Mid-tier Android devices (~$300) now get the single-core performance of a 2014 iPhone and the multi-core perf of a 2015 iPhone. Hardware Past As Performance Prologue. Mind The Gap.

Performance 124

Performance Network Cache Metrics 124

Brendan Gregg

AUGUST 8, 2017

They are demand on the system, albeit for software resources rather than hardware resources. ## Decomposing Linux load averages Can the Linux load average value be fully decomposed into components? Yes, I'd say so. Those threads are in the middle of doing work, and happen to block on a lock. They aren't idle.

Latency 111

Latency Metrics C++ Systems 111

Alex Russell

JANUARY 30, 2024

JavaScript-Heavy # Since at least 2015, building JavaScript-first websites has been a predictably terrible idea, yet most of the sites I trace on a daily basis remain mired in script. [1] This isn't charity; it's how teams ensure products stay functional, accessible, and reliable in a market awash in b t.

Performance 90

Performance Network Mobile Speed 90

Smashing Magazine

JANUARY 6, 2020

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). In real-life world, most products aren’t even close: an median bundle size today is around 417KB , which is up 42% compared to early 2015. Assets Optimizations.

Performance 145

Performance Cache Network Metrics 145

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). In real-life world, most products aren’t even close: an average bundle size today is around 400KB , which is up 35% compared to late 2015. We could also go beyond the bundle size budget though.

Performance 103

Performance Cache Metrics Network 103

Smashing Magazine

JANUARY 11, 2021

On the other hand, we have hardware constraints on memory and CPU due to JavaScript parsing and execution times (we’ll talk about them in detail later). compared to early 2015. In 2015, Google introduced Brotli , a new open-source lossless data format, which is now supported in all modern browsers. Assets Optimizations.

Performance 145

Performance Cache Media Metrics 145