https://www.gravatar.com/avatar/932f1b40c8d0202ce03a0df412bfb0ff?s=240&d=mp

chaomai's Odyssey

Notes of To Lock, Swap, or Elide: On the Interplay of Hardware Transactional Memory and Lock Free Indexing

解决了这么几个问题

  • Does HTM obviate the need for crafty lock-free index designs? No.

  • How does HTM differ from lock-free index designs?

  • Given that lock-free designs are still relevant, can HTM help simplify lock-free design techniques while maintaining good performance? Yes.

HTM

Transactional memory

ease the burden (a tedious and error prone task leading to deadlocks and race conditions) by delegating conflict detection and resolution from the developer to the system.

Hardware Transactional Memory

a best-effort model, optimistic concurrency.

Notes of CS100.1x Introduction to Big Data with Apache Spark (1)

The Big Data Problem

  • Growing data sources
  • Storage getting cheapper
  • But stalling CPU and storage bottlenecks

Hardware for Big Data

Problems with cheap hardware

  • Failures
  • Network
  • Uneven performance

What’s Hard About Cluster Computing

  • Divide work across machines
    • Must consider network, data locality
    • Moving data may be veay expensive
  • Deal with failures

Spark Essentials

PySpark

A Spark program consists of two programs, a driver program and a workers program.

Tmux

Basic

Creating Named Sessions

1
tmux new-session -s basic

or

1
tmux new -s basic

Detaching Sessions

CTRL-b, then d

Reattaching to Existing Sessions

1
tmux new-session -s basic

or

1
tmux ls
  1. If there is only session, then tmux attach.
  2. If there are multiple sessions, then tmux attach -t second_session. BTY, you can also use this when you only have one session.

Killing Sessions

  • In sesssion, use exit.
  • Kill specific session when not in session,
1
tmux kill-session -t basic

BTY, if there is no sessions, after typing tmux ls, you will get