aiter /* 有空再来改~ */

云计算

  • enterprise cloud (single organization)
  • public cloud (man organizations)

History

  • “cloud computing” appeared as eayly as 1996.
  • 2006 AWS的EC2(Elastic Compute Cloud)
  • 1960s, time-sharing(时分)

Types

IaaS (Infrastructure as a Service)

PaaS (Platform as a Service)

SaaS (Software as a Service)

Cloud Native

  • Containerization
  • CI/CD
  • Orchestration & Application definition
  • Observability & Analysis
  • Service proxy, Discovery & Mesh
  • Networking, Policy & Security
  • Distributed database & Storage
  • Streaming & Messaging
  • Container Registry & Runtime
  • Software distribution

The Twelve-Factor Application

  • 弹性伸缩
  • 集群中Node的伸缩
  • pod的伸缩

  • 可观测性

  • 消息队列
  • request/response messaging
  • request/reply
  • command
  • Event bus

Cloud-native data patterns

  • CAP
  • Consistency
  • Availability

  • Partition Tolerance

  • Relational database systems: CA(Consistency & Availability)

  • NoSQL databases: AP(abailability * Partition Tolerance)

  • Cloud-native resiliency
  • Monitoring and health
  • logging/monitoring/alerts
  • Logging with Elastic Stack

  • Monitoring in Azure Kubernetes Services(Prometheus)

  • Cloud-native identity/security
  • DevOps

cloud-native

2015-CHASE- Component High Availability-Aware Scheduler in Cloud Computing Environment

2013-Apache Hadoop YARN- Yet Another Resource Negotiator

  • Apache Hadoop
  • RM(Resource Manager)
  • tracks resource usage
  • node liveness
  • enforces allocation invariant
  • arbitrate contention
  • NM(Node manager)

  • AM(Application Manager)

  • YARN framework/application writers

  • YARN in the real-world
  • YARN at Yahoo!

  • MapReduce

  • availability-aware scheduling technique

Fuxi ParSync

  • distributed cluster scheduler
  • sharding & multiple schedulers
  • Workload statistics
  • 4 millon jobs & 4 billon tasks
  • 87% of tasks are completed in less than 10 seconds
  • Partitioned Synchronization

Omega: flexible, scalable schedulers for large compute clusters

  • Shared state

  • optimistic concurrency

  • 3.1 Monolithic schedulers
  • 3.2 Statically partitioned schedulers
  • 3.3 Two-level scheduling(Mesos) YARN
  • 3.4 Shared state(Omega)
  • fuxi is similar to Omega?
  • how to syn state
  • kubernetes’s scheduler
  • 异构资源 GPU

Benefits

  • 灵活 Flexible
  • 弹性 Scalable
  • 节约成本 Cost effective

CDN

  • Akamai
  • google CDN
  • microsoft azure CDN
  • AWS CDN

Compute

  • CPU
  • GPU
  • FPGA
  • TPU
  • ASIC

OpenStack

  • nova(Compute)
  • scheduler
    • Scheduler Evolution
    • Filter Scheduler
      • filtering:一个过滤器集合
      • weighting:排序
    • nova-scheduler是nova(compute)的一部分
    • nova使用scheduler来决定怎么分发(dispatch)资源的请求。VM在那一台物理机上运行。
    • 过滤器(filters)
      • 是否在指定的AZ、CPU/Memory/disk/architecture/hypervisor/vm mode/亲和性(Affinity)/反亲和性(AntiAffinity)
  • placement: 库存，提供API
    • resource provider inventories and usages. tracks the inventory and usage of each provider
    • compute node
    • shared storage pool
    • IP allocation pool
    • 有标准的DISK/MEMORY/VCPU，也可以自定义分类(custom resource classes )
    • nova-compute(注册、删除、管理) 和 nova-scheduler(调度)会和placement交互
  • scheduler提供调度的服务器和备选服务器 link
  • 组件(componets)关系
    • nova-api —-RPC—> super conductor
    • 还有cell conductors/cell-local conductors
    • super conductor 发送请求给 scheduler（不是cell conductors）
    • conductor 调度完成后，再发送生产请求到 nova-compute
    • 生产包含：网络port、磁盘、虚拟机 Walkthrough of a typical Nova boot request

组件关系

  • Horizon(Dashboard) / OpenAPI 对外界面
  • keystone 认证
  • nova 计算API、实例管理API （计算）
  • swift 对象存储 （存储-对象）
  • cinder 块存储 (存储-块)
  • glance 镜像服务 (镜像)
  • neutron 网络 (网络)
  • cells 可以部署更大的集群，每个cell是个独立的一组数据(相当于一个分片)。It supports very large deployments.

将集群中的机器分组，分组就是cell，

NP-hard

  • task scheduling
  • packing
  • placement

Heuristic techniques 启发式常用于一种近似方式解决优化问题。

  • is not guaranteed to be optimal、perfect、rational
  • immediate、short-term goal、approximation
  • derived from previous experiences with similar problems
  • readily accessible 易接近的
  • though loosely applicable 尽管松散的适用

computer CPU architecture

  • Branch predictor(分支预测) : 尝试猜测程序执行的逻辑，主要目的还是提速，让cpu的指令流水线(instruction pipeline)更快的执行
  • Static branch prediction。编译时就确定。
  • Dynamic branch prediction。运行时确定。
  • Random branch prediction
  • Next line prediction
  • One-level branch prediction

process schedule in OS

  • Long Term or job scheduler
  • Short term or CPU scheduler
  • Medium-term scheduler

cpu schedule in OS

  • Arrival Time: Time at which the process arrives in the ready queue.
  • Completion Time: Time at which process completes its execution.
  • Burst Time: Time required by a process for CPU execution.
  • Turn Around Time: Time Difference between completion time and arrival time.
  • Turn Around Time = Completion Time – Arrival Time
  • Waiting Time(W.T): Time Difference between turn around time and burst time.
  • Waiting Time = Turn Around Time – Burst Time

Objectives of Process Scheduling Algorithm

  • Max CPU utilization [Keep CPU as busy as possible]
  • Fair allocation of CPU.
  • Max throughput [Number of processes that complete their execution per time unit]
  • Min turn around time [Time taken by a process to finish execution]
  • Min waiting time [Time a process waits in ready queue]
  • Min response time [Time when a process produces first response]

Scheduling Algorithms

  • First Come First Serve (FCFS)
  • Shortest Job First (SJF)
  • Longest Job First (LJF)
  • Shortest Remaining Time First (SRTF)
  • Longest Remaining Time First (LRTF)
  • Round Robin Scheduling [the time-sharing system]
  • Priority Based scheduling (Non-Preemptive)
  • Highest Response Ratio Next (HRRN)
  • Multilevel Queue Scheduling

  • Multi level Feedback Queue Scheduling

  • kube-scheduler
    • List-Watch
    • scheduling
    • binding

Guidelines to follow when creating architectural diagrams

  • Keep structural and semantical consistency across diagrams/保持各图的结构和语义的一致性
  • Prevent diagrams fragmentation/防止图示碎片化
  • Keep traceability across diagrams/保持跨图表的可追溯性

  • Add legends next to architectural diagrams/在建筑图旁添加图例

  • Application architecture diagram
  • Integration architecture diagram
  • Deployment architecture diagram
  • DevOps architecture diagram
  • Data architecture diagram
  • shaps lines arrow
  • diagram legend

a picture is worth a thousand words

  • sh
  • bash
  • zsh
  • ksh

vlan & vxlan

Function compute

  • aws lambda
  • GCP Function

  • Azure Function

  • trigger

Containers at AWS

Gateway

  • ingress/egress
  • Higress

mesh

  • istio
  • envoy

nginx

  • upstream
  • proxy pass
  • http/3 QUIC

AWS

Fargate

gargate

Amazon Elastic Kubernetes Service (EKS)

eks

Amazon EventBridge

  • receive, filter, transform, route, and deliver events
  • 接受，过滤，转换，路由，投递

triggers

  • http
  • MQ

software life circle

  • life circle

  • ci/cd

  • dns

  • lvs
  • keeplived

service mesh

包粒度重传 3/4/7 gateway api gateway lvs

  • nginx:SSL/HTTP2/gRPC

  • prototype

  • cache
  • lru

CPU scheduler

  • CFS
  • period & quota
  • cpu.shares
  • affinity
  • isolation

References


blog comments powered by Disqus