/images/avatar.jpg
杂七杂八的,随手记录。

New Wine in an Old Bottle: Data-Aware Hash Functions for Bloom Filter

Table of Contents

本文为摘录(或转载),侵删,原文为: attachments/pdf/4/p1924-bhattacharya.pdf

FPR: Fault Positive Rate

Partitioned Bloom Filter ,

  • 方法

    • 将 BloomFilter 划分成 segments
    • 每个 segment 使用简单的、基于投影的哈希函数,
      • 函数通过数据计算得来
  • 效果

    • 减少误报 (两个数量级)
    • 相同误报率的话,提升 50% 的压缩率

1 INTRODUCTION

  • Projection Hash Bloom Filter (PHBF)

Performance Tip of the Week

本文为摘录(或转载),侵删,原文为: -https://abseil.io/fast/21 -https://abseil.io/fast/9 -https://abseil.io/fast

1 #9: Optimizations past their prime

1.1 Best practices

– 可能的情况下编写清晰、惯用的代码
Prefer writing clear, idiomatic code whenever possible. It is not only easier to read and debug, but in the long run, also easier for the compiler to optimize.

Prefix Filter: Better Than Bloom

Table of Contents

本文为摘录(或转载),侵删,原文为: ../pdf/d/p1311-even.pdf

1 Abstract

2 INTRODUCTION

  • What is a filter

    属于近似查询 (approximate membership query), 用于判断独具是否在集合中。

    • 如果输入在集合中,则必定输出 Yes
    • 如果不在集合中,则输出 Yes 的概率不能多于 \(\Sigma\)
  • What are filters used for
    • 与精确查询相比
      • 更省资源, 与数据量无关,而与 Σ (误报率)有关
      • 适于放在内存中,用于真正操作之前来过滤掉无效数据
        • 比如 join

Red Hat Enterprise Linux 7 Performance Tuning Guide

本文为摘录(或转载),侵删,原文为: attachments/pdf/c/red_hat_enterprise_linux-7-performance_tuning_guide-en-us.pdf

1 TOOLS

1.1 top

1.2 ps

1.3 vmstat – Virtual Memory Statistics

1.4 sar – System Activity Repoter

1.5 perf

1.6 tubostat

1.7 iostat

1.8 irqbalance

1.9 ss – socket statistics

1.10 numastat

1.11 numad