杂七杂八的，随手记录。

NBTree: a Lock-free PM-friendly Persistent B+-Tree for eADR-enabled PM Systems

yc 发布于 0001-01-01

本文为摘录(或转载)，侵删，原文为： attachments/pdf/a/p1187-zhang.pdf

PM stands for Persistent Memory，持久内存，它是一种非易失性内存，即使断电也能保留数据。
eADR 保证了 CPU 缓存中的数据在电源故障时会被刷新到 PM 中

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

yc 发布于 0001-01-01

Table of Contents

1 INTRODUCTION

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)

OceanBase 707M tpmC

yc 发布于 0001-01-01

Table of Contents

1 Introduction
2 DESIGN OVERVIEW
- 2.1 Goals
- 2.2 Criteria of Design
3 架构图
- 3.1 Application Layer
- 3.2 Proxy Layer
- 3.3 Data Service Layer
- 3.4 Zones
- 3.5 表
- 3.6 Node
- 3.7 SQL Engine
- 3.8 Multi-tenancy
- 3.9 Resource Isolation.
- 3.10 Features
4 STORAGE ENGINE
- 4.1 LSM Tree-Based Architecture
- 4.2 Asymmetric Read and Write
- 4.3 Daily Incremental Major Compaction
- 4.4 Replica Type
5 TRANSACTION PROCESSING ENGINE
- 5.1 Partition and Paxos Group
- 5.2 Timestamp Service
- 5.3 Transaction Processing
- 5.4 Isolation Level
- 5.5 Replicated Table
6 TPC-C BENCHMARK TEST
7 LESSONS IN BUILDING OCEANBASE
- 7.1 From NoSQL to NewSQL
- 7.2 Both cost and performance
- 7.3 Data validation
- 7.4 Partitioning vs. sharding

本文为摘录(或转载)，侵删，原文为： attachments/pdf/d/p3385-xu.pdf

Origami: A High-Performance Mergesort Framework

yc 发布于 0001-01-01

Table of Contents

1 ABSTRACT
2 INTRODUCTION
3 PIPELINE OVERVIEW
- 3.1 notation
- 3.2 better cache utilization,
- 3.3 4 Phases

本文为摘录(或转载)，侵删，原文为： attachments/pdf/c/p259-arman.pdf

1 ABSTRACT

Mergesort 优点：

不受数据倾斜影响
适于通过向量化执行来进行并行处理
适于多线程操作操作

Origami:

Performance of 10 G Ethernet Using Commodity Hardware

yc 发布于 0001-01-01

Table of Contents

1 INTRODUCTION
2 NETWORK PROCESSING IN THE LINUX KERNEL
- 2.1 Kernel Accounting
- 2.2 Packet Transmission 发送
- 2.3 Packet Reception 接收
- 2.4 Incorrect SoftIRQ Accounting
3 TESTBED AND KERNEL CONFIGURATION
4 UDP TRANSMISSION
5 TCP TRANSMISSION
6 TCP TRANSMISSION WITH ZERO-COPY
7 THE HARDWARE OFFLOAD FUNCTIONALITIES OF THE NETWORK ADAPTER
8 FIBRE CHANNEL TO 10 GIGABIT ETHERNET TESTS
9 CONCLUSIONS

本文为摘录(或转载)，侵删，原文为： attachments/pdf/3/performanceOf10GbE.pdf

Performance Tip of the Week

yc 发布于 0001-01-01

Table of Contents

1 #9: Optimizations past their prime
- 1.1 Best practices
- 1.2 #21: Improving the efficiency of your regular expressions

本文为摘录(或转载)，侵删，原文为： 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.