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

Scalegrid

FEBRUARY 8, 2024

A distributed storage system is foundational in today’s data-driven landscape, ensuring data spread over multiple servers is reliable, accessible, and manageable. This guide delves into how these systems work, the challenges they solve, and their essential role in businesses and technology.

Storage 130

Storage Systems Big Data Azure 130

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

HammerDB

OCTOBER 23, 2023

Stored Procedures and Client SQL comparison To test the stored procedures and client implementations, we ran both workloads against a system equipped with Intel Xeon 8280L. The data shows a scripted automated workload running a number of back to back tests each time with an increasing number of virtual users. On MySQL, we saw a 1.5X

Benchmarking 62

Benchmarking Database Network C++ 62

VoltDB

JUNE 1, 2017

It was – like the hypothetical movie I describe above – more than a little bit odd, as you could leave a session discussing ever more abstract layers of virtualization and walk into one where they emphasized the critical importance of pinning a network interface to a specific VM for optimal performance.

Hardware 56

Hardware Latency Virtualization Cloud 56

Brendan Gregg

AUGUST 18, 2019

For example, iostat(1), or a monitoring agent, may tell you your average disk latency, but not the distribution of this latency. For smaller environments, it can be of more use helping eliminate latency outliers. bpftrace uses BPF (Berkeley Packet Filter), an in-kernel execution engine that processes a virtual instruction set.

Latency 68

Latency C++ Cache Programming 68

Dotcom-Montior

NOVEMBER 11, 2020

Serverless computing can be a huge benefit to organizations that don’t have the necessary resources or teams to manage physical resources, like servers/hardware, and all the maintenance and licensing that goes along with that, allowing them to focus on developing their code and applications. Benefits of a Serverless Model.

Serverless 57

Serverless Monitoring Lambda Latency 57

VoltDB

JUNE 1, 2017

It was – like the hypothetical movie I describe above – more than a little bit odd, as you could leave a session discussing ever more abstract layers of virtualization and walk into one where they emphasized the critical importance of pinning a network interface to a specific VM for optimal performance.

Hardware 40

Hardware Latency Virtualization Cloud 40

Dotcom-Montior

DECEMBER 8, 2021

Systems, web applications, servers, devices, etc., SREs and DevOps teams can use these incidents to build back better and improve their systems and services. Now that we have talked about what an incident is, incident management is the process by which teams resolve these events and bring systems and services back to normal operation.

Social Media 90

Social Media Monitoring Latency DevOps 90

All Things Distributed

JANUARY 23, 2012

Werner Vogels weblog on building scalable and robust distributed systems. The Amazon Virtual Private Cloud extends on-premises compute with all the power of AWS, making it elastic, scalable and highly reliable. All Things Distributed. Expanding the Cloud - The AWS Storage Gateway. By Werner Vogels on 23 January 2012 12:01 AM.

Storage 116

Storage AWS Virtualization Cloud 116

Abhishek Tiwari

JULY 12, 2018

In a nutshell, a data pipeline is a distributed system. A data pipeline is a software which runs on hardware. The software is error-prone and hardware failures are inevitable. Every time data spikes - a phenomenon which will be not predictable in most cases - overall latency to process the data using data pipeline will go up.

Latency 63

Latency Analytics Scalability Engineering 63

HammerDB

DECEMBER 5, 2020

A full understanding of why this is important requires some knowledge of the evolution of database hardware and software. For TPC-C this meant enough available spindles to reduce I/O latency and for TPC-H enough bandwidth for data throughput. I.e. if system A generated 1.5X XML Connect Pooling (test distributed clusters).

C++ 40

C++ Benchmarking Virtualization Open Source 40

Dynatrace

JANUARY 26, 2021

Traditional computing models rely on virtual or physical machines, where each instance includes a complete operating system, CPU cycles, and memory. VMware commercialized the idea of virtual machines, and cloud providers embraced the same concept with services like Amazon EC2, Google Compute, and Azure virtual machines.

Serverless 245

Serverless Efficiency Lambda Azure 245

Abhishek Tiwari

DECEMBER 27, 2017

Containerized data workloads running on Kubernetes offer several advantages over traditional virtual machine/bare metal based data workloads including but not limited to. direct access to raw block storage [18] without the abstraction of a filesystem for workloads that require consistent I/O performance and low latency.

Big Data 63

Big Data Storage Benchmarking Hardware 63

Dotcom-Montior

NOVEMBER 24, 2021

Web developers or administrators did not have to worry or even consider the complexity of distributed systems of today. Great, your system was ready to be deployed. Once the system was deployed, to ensure everything was running smoothly, it only took a couple of simple checks to verify. What is a Distributed System?

Systems 74

Systems Monitoring Hardware Network 74

Brendan Gregg

SEPTEMBER 25, 2021

A Cassandra database cluster had switched to Ubuntu and noticed write latency increased by over 30%. CLI tools The Cassandra systems were EC2 virtual machine (Xen) instances. As a Xen guest, this profile was gathered using perf(1) and the kernel's software cpu-clock soft interrupts, not the hardware NMI.

Speed 126

Speed Java AWS Virtualization 126

All Things Distributed

MARCH 11, 2016

Build evolvable systems. But we couldn’t adopt the old style approach of upgrading systems through a maintenance outage, as many businesses around the world are relying on our platform for 24/7 availability. This is a given, whether you are using the highest quality hardware or lowest cost components. Primitives not frameworks.

AWS 147

AWS Hardware Retail Virtualization 147

The Morning Paper

DECEMBER 10, 2019

This is a companion paper to the " persistent problem " piece that we looked at earlier this week, going a little deeper into the object pointer representation choices and the mapping of a virtual object space into physical address spaces. Both abstractions must be implemented in a way that is efficient using existing hardware.

Hardware 52

Hardware Virtualization Operating System Programming 52

Brendan Gregg

SEPTEMBER 25, 2021

A Cassandra database cluster had switched to Ubuntu and noticed write latency increased by over 30%. CLI tools The Cassandra systems were EC2 virtual machine (Xen) instances. As a Xen guest, this profile was gathered using perf(1) and the kernel's software cpu-clock soft interrupts, not the hardware NMI.

Speed 52

Speed Java AWS Virtualization 52

The Morning Paper

DECEMBER 8, 2019

At the start of November I was privileged to attend HPTS (the High Performance Transaction Systems) conference in Asilomar. Byte-addressable non-volatile memory,) NVM will fundamentally change the way hardware interacts, the way operating systems are designed, and the way applications operate on data. PLOS’19.

Hardware 56

Hardware Programming Media Storage 56

Brendan Gregg

SEPTEMBER 25, 2021

A Cassandra database cluster had switched to Ubuntu and noticed write latency increased by over 30%. CLI tools The Cassandra systems were EC2 virtual machine (Xen) instances. As a Xen guest, this profile was gathered using perf(1) and the kernel's software cpu-clock soft interrupts, not the hardware NMI.

Speed 40

Speed Java AWS Virtualization 40

HammerDB

OCTOBER 18, 2019

So lets take an Ubuntu system with Platinum 8280 CPUs with the following Ubuntu OS, reboot and check the CPU configuration before running any tests. So by default, the system boots into powersave. hardware limits: 1000 MHz - 4.00 hardware limits: 1000 MHz - 4.00 uname -a Linux ubuntu19 5.3.0-rc3-custom 3834 2696 .

Benchmarking 63

Benchmarking Performance Hardware C++ 63

The Morning Paper

APRIL 14, 2019

Just as today’s systems offer memory protection, they call this time protection. The paper sets out what we can do in software given today’s hardware, and along the way also highlights areas where cooperation from hardware will be needed in the future. set out a principled basis for protecting against this class of attacks.

Hardware 44

Hardware Cache Latency Speed 44

Percona

APRIL 17, 2023

In this blog post, we will discuss the best practices on the MongoDB ecosystem applied at the Operating System (OS) and MongoDB levels. Operating System (OS) settings Swappiness Swappiness is a Linux kernel setting that influences the behavior of the Virtual Memory manager when it needs to allocate a swap, ranging from 0-100.

Best Practices 52

Best Practices Design Tuning Database 52

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. Media Session API. Offscreen Canvas.

Media 145

Media Games Education Engineering 145

Brendan Gregg

AUGUST 8, 2017

Linux load averages are "system load averages" that show the running thread (task) demand on the system as an average number of running plus waiting threads. This measures demand, which can be greater than what the system is currently processing. then your system is idle. - cat /proc/loadavg. 42/3411 43603. 42/3411 43603.

Latency 111

Latency Metrics C++ Systems 111

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. System Setup: CPU, Memory and I/O Configuration. hardware limits: 1000 MHz - 3.80

Best Practices 40

Best Practices Scalability Performance C++ 40

Brendan Gregg

SEPTEMBER 5, 2017

What follows are topics that may be of interest to anyone looking to migrate their systems and skillset: scan these to find topics that interest you. ## ZFS ZFS is available for Linux via the [zfsonlinux] and [OpenZFS] projects, and more recently was included in Canonical's Ubuntu Linux distribution: Ubuntu Xenial 16.04 LTS (April 2016).

Virtualization 76

Virtualization AWS Engineering Hardware 76

All Things Distributed

JANUARY 18, 2012

Werner Vogels weblog on building scalable and robust distributed systems. The original Dynamo design was based on a core set of strong distributed systems principles resulting in an ultra-scalable and highly reliable database system. Amazon DynamoDB offers low, predictable latencies at any scale. All Things Distributed.

Scalability 167

Scalability Database Ecommerce Latency 167

SQL Server According to Bob

JANUARY 11, 2020

This will help you increase system performance and avoid I/O environment errors. Important When planning, deploying, and maintaining a Microsoft SQL Server installation, ensure that the I/O system supports all the factors outlined in this article. The modifications persist even in the event of a system failure.

Servers 40

Servers Cache Media Hardware 40

SQL Server According to Bob

JANUARY 11, 2020

​​ Without limiting the rights under copyright, no part of this document may be reproduced, stored in or introduced into a retrieval system, or transmitted in any form or by any means (electronic, mechanical, photocopying, recording, or otherwise), or for any purpose, without the express written permission of Microsoft Corporation.

Servers 40

Servers Cache Database Media 40

Dynatrace

AUGUST 5, 2019

Lift & Shift is where you basically just move physical or virtual hosts to the cloud – essentially you just run your host on somebody else’s hardware. Remember: This is a critical aspect as you do not want to migrate a service and suddenly introduce high latency or costs to a system that you forgot about having a dependency with!

Cloud 31

Cloud Traffic Database Network 31

Smashing Magazine

JANUARY 11, 2021

Estimated Input Latency tells us if we are hitting that threshold, and ideally, it should be below 50ms. Designed for the modern web, it responds to actual congestion, rather than packet loss like TCP does, it is significantly faster , with higher throughput and lower latency — and the algorithm works differently.

Performance 145

Performance Cache Media Metrics 145