Reading Critiques

  1. Addressing followings:
    • Summary of major innovations
    • What the problems the paper mentioned
    • How about the important related work/papers
    • What are some intriguing aspects of the paper
    • How to test/compare/analyze the results
    • How can the research improved
    • If you write this paper, then how would you do?
    • Results from this paper
    • Give the survey paper list in the same research area
  2. One Project

importance knobs how to add compared knobs or not adding new.

Rethinking OS Components

Single-machine OS/VMM:

  1. Strategy

    • how to organize OS for development/evolution/performance/security

    • how to use multi-processor machines

  2. Components

    • Mem management
    • Process management
    • Synchronization
    • File systems & device support

Mem management:

  1. Purposes
    • Virtual memory
    • Swapping: moves #p to disks
    • Paging: #p runs only the active pages in mem
    • Buffer Cache
  2. Strategy
    • how to share mem with multi-machines
    • how to simplify mem management as mem becomes abundant

Process management:

  1. Purposes
    • 多 #p 并发执行
    • context switching: Interrupt
  2. Strategy
    • Achive fairness/throughput/responsiveness
    • Reduce/avoid the cost

Synchronization:

  1. Purposes
    • 联系 #p，如在 context switching 阶段需要coordinating threads
  2. Strategy
    • Atomic actions
    • Mutual exclusion：critical region一个时刻一个 thread 执行
    • Semephores: atomic, counter-based locks
    • Avoid Deadlock

File Systems:

  1. Purposes
    • File: data+attributes
    • File services. E.g. Naming, Access, chmod
  2. Strategy
    • 多机间共同运行 File systems
    • 多机间 copy 时 provide consistency/availability/reliability
    • How to handle large data sets

I/O systems:

  1. Purposes:

    • 速度慢于memory
    • Caching

  2. Strategy：

    1. 多机间能有共同 share 的 resources 来提高性能
    2. handle new devices with new characteristics

    I’d like to challenge this!

Performance

  • 如何利用 Multi-core 性能: E.g. AMD 64cores & Intel 48 cores

  • Benchmark: Speedup / throughout /

Kernel synchronizations protecting the shared data are the main bottlenecks on multi-core platform

Reliability

  • Includes availability/security/fault tolerance/durability

  • Make sure data services when system crashes, disk crashed , etc

Threat Analysis

  1. Protect what, why?
  2. What are the vulnerabilities of those assets
  3. Who might exploit vulnerabilities
  4. How to prevent a specific threat
  5. Consider cost

Correctness & Summary