Brendan Gregg

APRIL 14, 2022

offer letter logo (2014) flame graphs (2014) eBPF tools (2014-2019) PMC analysis (2017) my pandemic-abandoned desk (2020); office wall I joined Netflix in 2014, a company at the forefront of cloud computing with an attractive work culture. Netflix has been the best job of my career so far, and I'll miss my colleagues and the culture.

Java 52

Java Cache Cloud AWS 52

Alex Russell

JUNE 19, 2020

Perhaps there's a better solution, but I've been iterating on content and templates by firing up a linux terminal on my 2017 Pixelbook and starting the built-in 11ty filewatcher and browser sync tools. Eleventy is I/O heavy, and for correctness sake, it didn't do much caching. Copied 1189 files / Wrote 1175 files in 5.75 MiB | 2.99

Cache 49

Cache Hardware Speed Cloud 49

Brendan Gregg

SEPTEMBER 5, 2021

I summarized this case study at [Kernel Recipes] in 2017; it is an old story that's worth resharing here. ## 1. The ZFS Adapative Replacement Cache (ARC) is the main memory cache for the file system. The ARC contains lists of cached buffers for different memory types. I worked on this code back at Sun.

Cache 40

Cache Cloud Java Tuning 40

KeyCDN

JANUARY 31, 2019

As for Romania, it covers an area of over 238,000 square kilometers and as of 2017 had a population of almost 20 million. In 2017, the average speed was 21.33Mb/s ranking them 18th in the world. As of 2016, Bucharest alone had a population of over 1.83 million people. million Romanian eCommerce users who spend an average of $237.38

Internet 48

Internet Internet Ecommerce Speed 48

Adrian Cockcroft

JANUARY 29, 2023

Bill Kaiser of NewRelic published this blog in 2017 which goes some way towards what I’m talking about, but since then I have figured out a new way to interpret the data. Changes in behavior of the system from minute to minute is going to change the height of each peak, as the workload mix and cache hit rates change.

Lambda 98

Lambda Latency Cache C++ 98

MachMetrics

DECEMBER 14, 2019

In 2017, mobile internet usage passed desktop as the majority. It is typically reduced via server-side optimizations, such as enabling caching and database indexes. What is the number of resources I should stay under? How fast is the average time to first byte (server delay)? What impact does the increased use of mobile devices have?

Best Practices 105

Best Practices Speed Mobile Google 105

SQL Performance

SEPTEMBER 17, 2019

For years SSRS was bundled with the installation of SQL Server, which helped add to some of the confusion around licensing and support for the product, so beginning with SSRS 2017, the installation package for Reporting Services is a separate download. Reporting Services Infrastructure.

Tuning 67

Tuning Servers Database Best Practices 67

KeyCDN

OCTOBER 24, 2018

According to Wikipedia , the average connection speed in Q1 of 2017 was 20.5Mbps. This is a massive increase from year 2000 when only 37.2% of the population were Internet users. In terms of Internet speed, Finland ranks as one of the fastest countries in the world. This puts them in 6th place for fastest connection speeds.

Internet 48

Internet Internet Speed Network 48

ACM Sigarch

FEBRUARY 23, 2018

Richard Thaler won the 2017 Nobel prize in Economics for his work on Behavioral Economics. Architects were always aware of the impact speculative execution has on the performance of microarchitectural structures from branch predictors to caches. However, we did not adequately consider speculative execution’s impact on security.

Architecture 47

Architecture Cache Hardware Virtualization 47

Percona

JUNE 4, 2023

This blog post was originally published in November 2017 and was updated in June 2023. 3 GA 23 May 2017 10.2.6 Some features, like time-delayed replication that were present in MySQL since 2013, only make an appearance in MariaDB Server in 2017! MySQL Percona Server for MySQL MariaDB Server 3 December 2010 5.5.8

Open Source 52

Open Source Servers Storage Scalability 52

SQL Server According to Bob

DECEMBER 18, 2018

This means data can be stored in file system cache, non-stable media. ) The issue, as described in the link, is that the sync returns the error but may clear the state of the cached pages. The next sync returns ESUCCESS, meaning the the write(s), which were in cache, do not flush to stable media but applications were told they did.

Servers 40

Servers Media Cache Software Engineering 40

SQL Server According to Bob

DECEMBER 18, 2018

Durability is a cornerstone of any database system and starting with SQL Server 2017 on Linux Cumulative Update 6 (CU6), SQL Server on Linux enables “Forced Flush” behavior as described in this article , improving durability on non-Fua optimized systems. “Be Linux open command flag used to bypass file system cache. SQL Server.

Servers 90

Servers Media Cache Storage 90

SQL Performance

JUNE 5, 2019

Subsequent executions have the same plan (as that’s what was cached) and the same memory grant, but we get a clue it’s not enough because there’s a sort warning. While the plan in the plan cache has memory grant information updated when memory feedback occurs, this information does not get applied to the existing plan in Query Store.

Servers 41

Servers Cache Database Metrics 41

High Scalability

NOVEMBER 30, 2018

8 : successful Mars landings; $250,000 : proposed price for Facebook Graph API; 33 : countries where mobile internet is faster than WiFi; 1000s : Facebook cache poisoning; 8.2 during the 2017-18 academic year, on top of a 3.3% during the 2017-18 academic year, on top of a 3.3% They'll love it and you'll be their hero forever.

Internet 28

Internet Internet Scalability Blockchain 28

The Netflix TechBlog

JULY 1, 2019

Our UI runs on top of a custom rendering engine which uses what we call a “surface cache” to optimize our use of graphics memory. Surface Cache Surface cache is a reserved pool in main memory (or separate graphics memory on a minority of systems) that the Netflix app uses for storing textures (decoded images and cached resources).

Cache 166

Cache Design Testing Development 166

The Symphonia

MAY 22, 2018

Gojko Adzic has done some great speaking and writing on his experience here, and I included the link to his talk from late 2017. While this idea exists in the Microservices world too, the extent to which that can be taken, with great effect, with Serverless is astonishing to me. I rewrote the “ State ” section?—?I

Serverless 40

Serverless Architecture Lambda Azure 40

CSS - Tricks

APRIL 1, 2021

A 2017 study by Akamai says as much when it found that even a 100ms delay in page load can decrease conversions by 7% and lose 1% of their sales for every 100ms it takes for their site to load which, at the time of the study, was equivalent to $1.6 Compressing, minifying and caching assets. The final thing we can check is caching.

Cache 92

Cache Social Media Media Website 92

MachMetrics

JUNE 30, 2021

Daniel An, Google, 2017. According to Google SOASTA research, 2017 the probability of page bounce rate increases with increasing page loading time, because it greatly improves the user experience. Once you have configured the cache settings properly here comes the help of server response time and TTFB (Time To First Byte).

Cache 78

Cache Speed Google Website 78

Alex Russell

MARCH 6, 2021

TL;DR: A lot has changed since 2017 when we last estimated a global baseline resource per-page resource budget of 130-170KiB. A then-representative $200USD device had 4-8 slow (in-order, low-cache) cores, ~2GiB of RAM, and relatively slow MLC NAND flash storage. The Moto G4 , for example. Here begins our 2021 adventure. Hard Reset.

Performance 124

Performance Network Cache Metrics 124

Alex Russell

DECEMBER 18, 2022

The talks are up on YouTube , including my own (behind a paywall, but the slides are freely available [1] ): The talk, like this post, is an update on network and CPU realities this series has documented since 2017. These are depressingly similar specs to devices I recommended for testing in 2017. 2023 Content Targets #. 4GiB of RAM.

Performance 113

Performance Network Mobile Latency 113

Percona

JANUARY 17, 2019

Aurora Parallel Query response time (for queries which can not use indexes) can be 5x-10x better compared to the non-parallel fully cached operations. Here is a summary of the queries: Simple queries: select count(*) from ontime where flightdate > '2017-01-01'. The second and third run used the cached data.

Cache 47

Cache C++ AWS Airlines 47

SQL Server According to Bob

MAY 19, 2019

The reduction effort started when we shipped SQL Server 2017 ( [link].) The lock manager has partitions, lock block cache and other structures. Reduce the number of partitions and size of the cache. IO Request Caches. SQLPAL may cache I/O request structures with each thread. SQL Server 2017. Lock Manager.

IoT 61

IoT Servers Development Cache 61

Smashing Magazine

JULY 29, 2019

MB , that suggests I’ve got around 29 pages in my budget, although probably a few more than that if I’m able to stay on the same sites and leverage browser caching. There’s a trade-off to be made here, as external stylesheets can be cached but inline ones cannot (unless you get clever with JavaScript ). Let’s talk about caching.

Cache 96

Cache Google Mobile Network 96

Sutter's Mill

FEBRUARY 13, 2017

Next, I’ll be heading to Bristol to catch the end of the ACCU 2017 conference, and give the closing talk on “Something(s) New in C++.” Here is the current abstract: By the time the ACCU 2017 conference begins, C++17 is expected to be technically complete and in its final approval ballot. What comes next?

Latency 51

Latency C++ Hardware Performance 51

Alex Russell

APRIL 29, 2021

WebGL 2 launched for other platforms on Chrome and Firefox in 2017. An extension to Service Workers that enables browsers to present users with cached content when offline. A-series chips have run circles around other ARM parts for more than half a decade, largely through gobsmacking amounts of L2/L3 cache per core.

Media 145

Media Games Education Engineering 145

All Things Distributed

OCTOBER 2, 2017

In support of Amazon Prime Day 2017, the biggest day in Amazon retail history, DynamoDB served over 12.9 In addition, DynamoDB Accelerator (DAX) a fully managed, highly available, in-memory cache further speeds up DynamoDB response times from milliseconds to microseconds and can continue to do so at millions of requests per second.

Internet 128

Internet Internet AWS Performance 128

Smashing Magazine

FEBRUARY 3, 2022

These numbers are based on first-page load — caching seems very efficient for subsequent page loads. This post from 2017 outlines some stats and some possible reasons.) Two embeds. Description. JS (uncached). Single embed. Two embeds. The consensus seems to be that users are far more likely to share content in other ways (i.e.

Social Media 101

Social Media Media Website Airlines 101

SQL Performance

NOVEMBER 4, 2021

First introduced in SQL Server 2017 Enterprise Edition, an adaptive join enables a runtime transition from a batch mode hash join to a row mode correlated nested loops indexed join (apply) at runtime. For brevity, I’ll refer to a “correlated nested loops indexed join” as an apply throughout the rest of this article.

Servers 98

Servers Design Database Code 98

SQL Performance

APRIL 8, 2019

Back on December 5, 2017, Microsoft announced that they were using AMD EPYC 7551 processors in their storage-optimized Lv2-Series virtual machines. The L3 cache size is 64MB. The L3 cache size is 64MB. Since then, Microsoft has changed the naming of this series to Lsv2. lanes for I/O connectivity. Azure Lsv2 Details.

Azure 42

Azure Virtualization Storage Benchmarking 42

Alex Russell

OCTOBER 22, 2017

— Alex Russell (@slightlylate) October 4, 2017. Add onto that the yawning chasm between low-end and high-end device performance thanks to chip design factors like cache sizes, and it can be difficult to know where to set a device baseline. We now have everything we need to create a ballpark perf budget for a product in 2017.

Performance 68

Performance Benchmarking Network Mobile 68

SQL Performance

JULY 16, 2019

They will also have up to 256MB of L3 cache per processor. These include your overall server CPU capacity, your single-threaded CPU performance, your memory density and capacity, your total I/O capacity, and your SQL Server 2017/2019 license costs. There are many reasons!

Servers 46

Servers Hardware Architecture Retail 46

Enprowess

JANUARY 12, 2022

So it is convenient for all to use irrespective of internet speed and it works offline using cached data. Twitter Lite PWA : It was integrated as the standard User Interface in 2017. Offline Support : SPA consumes less bandwidth; meanwhile, it loads pages once only. Pinterest PWA: Pins increased by 401%. AI-powered Chatbots.

Blockchain 59

Blockchain IoT Development Serverless 59

SQL Performance

AUGUST 23, 2019

The details given here apply to SQL Server 2017 — the most recently released version at the time of writing. The SQL Server batch mode implementation of a Bloom filter is optimized for modern CPU cache architectures and is known internally as a complex bitmap. Features specific to earlier versions will mentioned inline as we go along.

Servers 75

Servers Engineering Efficiency Retail 75

SQL Server According to Bob

DECEMBER 31, 2018

. // Do we need to start the read-ahead to suck data into file system cache if (iCookie > -1) { tReadAhead.Start(iCookie); } }. iBytesRead = s.Read(bData, 0, c_iMaxReadSize); } }. . // Do we need to start the read-ahead to suck data into file system cache if (iCookie > -1) { tReadAhead.Start(iCookie); } }.

Efficiency 40

Efficiency Code Servers Cache 40

Smashing Magazine

JANUARY 21, 2021

After the latest redesign in late 2017, it was Ilya Pukhalski on the JavaScript side of things (part-time), Michael Riethmueller on the CSS side of things (a few hours a week), and yours truly, playing mind games with critical CSS and trying to juggle a few too many things. Moving From Automated Critical CSS Back To Manual Critical CSS.

Performance 145

Performance Cache Mobile Network 145

Smashing Magazine

JANUARY 21, 2019

Native Client-Side Cache/Data Store. For instance, the returned data for each request can be added into a client-side cache containing all data requested by the user throughout the session. And just by iterating all the values under modulesettings["share-on-social-media"].modules Extensibility And Re-purposing.

Social Media 64

Social Media Database Media Website 64

Smashing Magazine

JANUARY 7, 2019

The idea is quite straightforward: Push the minimal code needed to get interactive for the initial route to render quickly, then use service worker for caching and pre-caching resources and then lazy-load routes that you need, asynchronously. An application shell is the minimal HTML, CSS, and JavaScript powering a user interface.

Performance 103

Performance Cache Metrics Network 103