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. BBR: Congestion-Based Congestion Control,” [link] 2016 - [Gregg 16] Brendan Gregg, “Unikernel Profiling: Flame Graphs from dom0,” [link] Jan 2016 - [Gregg 16b] Brendan Gregg, “Linux 4.X Ford, et al., “TCP

Performance 52

Performance Latency Hardware Storage 52

Brendan Gregg

SEPTEMBER 5, 2017

LTS (April 2016). I wrote about it in a previous post, [DTrace for Linux 2016]. Here's some output from my zfsdist tool, in bcc/BPF, which measures ZFS latency as a histogram on Linux: # zfsdist. Tracing ZFS operation latency. The hardest part on Linux is now done: kernel support. Hit Ctrl-C to end. ^C

Virtualization 75

Virtualization AWS Engineering Hardware 75

Alex Russell

APRIL 29, 2021

At the time of the last Confluence run, the gap had stretched to nearly 1000 APIs, doubling since 2016. For heavily latency-sensitive use-cases like WebXR, this is a critical component in delivering a good experience. is access to hardware devices. Some commenters appear to confuse unlike hardware for differences in software.

Media 145

Media Games Education Engineering 145

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

O'Reilly Software

APRIL 4, 2018

A message-oriented implementation requires an efficient messaging backbone that facilitates the exchange of data in a reliable and secure way with the lowest latency possible. In 2016, Apache Spark introduced Structured Streaming , a new streaming engine based on the SparkSQL abstractions and runtime optimizations.

Architecture 80

Architecture Scalability Google Operating System 80

Alex Russell

OCTOBER 22, 2017

It simulates a link with a 400ms RTT and 400-600Kbps of throughput (plus latency variability and simulated packet loss). Simulated packet loss and variable latency, however, can make benchmarking extremely difficult and slow. Our baseline, then, should probably trade lower throughput/higher-latency for packet loss.

Performance 74

Performance Network Benchmarking Mobile 74

John McCalpin

NOVEMBER 5, 2016

In the latest (October 2016) revision of Intel’s Instruction Extensions Programming Reference , Intel has disclosed a fairly dramatic departure from these “traditional” approaches. With 2 FMA units that have 5-cycle latency, the code must implement at least 2*5=10 independent accumulators in order to avoid stalls.

Cache 40

Cache C++ Latency Hardware 40