# SLA and Virtualization

## SLA

### Definition

- Service Level Agreement: a service-based legal agreement between a
  **provider** and a **subscriber**
- Accepted before a cloud service can be used
- Content
    - Provider's responsibility
    - Subscriber's responsibility, and entitlements
    - Billing and accounting system

### Types

- Service based SLA:
    - Offer shared resources, used by many customers
    - Offered to cloud users
- Customer based SLA
    - Agreement between individual customer or group, to provide non-shared
      **specific** services
- Multi-level SLA:
    - Corporate level: generic and cover all user of an organization
    - Customer level: specific group of user
    - Service: specific type of service

### Boundaries

- Organizational boundary
    - Physical perimeter, owned and governed by organization or cloud provider
- Trust boundary
    - When cloud user extend trust beyond its physical boundary, and include
      resources from the cloud

### Characteristics of cloud

- On-demand usage: use more cloud resources only when they need it
- Multi-tenancy aka. Resource pooling: Resources are **instances** that can be
  used by many users
- Scalability: Able to automatically scale up or down based on the settings and
  agreements
- Ubiquitous access: Can be accessed everywhere with lots of devices
- Measured usage: keep track and record the amount of IT resources used by
  consumers
    - Used for accounting and auditing
- Resilience: distribute and replicate resources to provide redundancy, and fail
  over mechanism

### QoS

- SLA is based on **Quality of Service**
- Combination of metrics relating to requirements, and satisfaction of services
  offered by a provider.
- Quantitative or qualitative
- Parameters:
    - **Reliability** availability
    - **Security** and trust
    - **Scalability**
    - **Data** integrity and data protection
    - **Resilience**
    - Data **recovery** time
    - Support

## Virtualization

### Definition

- The process of creating the **virtual** version of hardware, database,
  operating system and application from real copies, using virtual machine
  **monitor** to emulate the physical functionalities.
- Multiple virtual machines are **multiplexed** in to the same hardware, to
  serve as an abstraction of physical OS
- Guest OS is not bound to the physical hardware.
- **Guest** OS runs on **host** OS, and they can be different

### Levels of virtualization

- Instruction Set Architecture level: ISA **emulation**, which is used to run
  many legacy codes written for a different hardware configuration, like
  _Dynamo_
- Hardware level: VMs are located directly on top of **hardware**, allows the
  virtualization of each of the hardware components, which could be the
  input-output device, the memory, the processor, etc, like _VMWare_
- Operating system level: an operating system (OS) virtualization paradigm in
  which the kernel allows the existence of multiple isolated user space
  instances, like _Jails_
- Library level: translate API calls to user libraries, like _vCUDA_
- Application level: allows high-level language programs to be compiled easily.
  java's _jvm_

### Types of Virtualizations

- Desktop: Create desktop computer from physical desktop machine
- Hardware: Virtual computer with OS that represents and act as a real physical
  computer.
- Nested virtualization: Virtual machine inside another one
- CPU virtualization: Virtual CPU are created to reduce stress on host
- Memory virtualization: virtual memory is created and run on the VM, controlled
  by the virtual CPU of VM, but can't access memory address outside virtual
- IO virtualization: Route IO request between virtual devices and physical
  hardware

### Virtual clusters

- Physical cluster: interconnected by physical network
    - Failure may cause a falling physical node
- Virtual cluster: VMs intereconnected by a virtual network across several
  physical networks
    - Node can be either physical or virtual
    - Cluster VMs can run different OS
    - To enhance **utilization** and **flexibility**
    - Can be freely replicated, and is **fault** tolerant easy to recover
    - Nodes can grow or shrink
    - Failure won't cause shutdown to host machine

### Migration of VM

#### Introduction

- VM can be moved from one cluster to another, which **isn't possible** with
  physical hosts.
- Four ways:
    - Guest-based manager, where cluster manager is in guest vm
    - Host based manager: supervises the guest system, and can restart guest
      system remotely
    - Independent cluster manager on both host and guest
    - Use integrated cluster manager

#### Six (five) steps of VM migration

- Start migration (for some reason this is counted as step 0 and 1???)
    - Determine VM to migrate
    - Determine destination host
- Transfer memory
    - State is stored in memory, so transfer memory first
- Suspend VM and copy remaining data
    - Execution is suspended when last memory is transfered.
    - Non memory components like network and cpu are sent to destination
- Commit
    - After all data, VM reloads and continue execution
- Activate
    - Network is redirected to new VM
    - remove the original VM from source host

### Mechanism for cloud infra: building blocks

- Environment:
    - Logical network perimeter: Virtual network boundary, encompass and isolate
      a group of resources, that may be **physically** distributed., to _isolate
      cloud resource from unauthorized users_
    - Virtual Server: behaves exactly like a physical one, and running on the
      same physical server which is shared by multiple cloud users
    - Resource replication: create multiple of same resource to various
      locations, physically distributed around the world.
    - Ready made environment: provided by PaaS (platform as a service)
    - Failover system: use clustering to provide redundancy
    - Hypervisor: generate virtual server instances
    - Multi device broker: converts and exchanges information
    - Resource cluster: cloud resources are geographically distributed and group
      resources and instances together
    - State management database, used to temporarily maintain state data for
      software programs to improve performance.
- Scaling
    - Automated Scaling listener: monitors communication between consumers and
      services for dynamic scaling up or down.
    - Load balancer: cloud runtime mechanism that distributes workload: data
      processing, computation, network traffic across two or more cloud
      resources, for better performance( automated horizontal scaling)
- Monitoring
    - Cloud usage monitor: Autonomous software used for collecting and
      processing cloud resource usage data.
    - SLA monitor: monitors the performance to ensure they are fulfilling the
      Quality of Service requirement
    - Pay per use monitor: monitor the usage and calculate the price according
      to the agreement
    - Audit monitor: track data for network, to confirm to regulatory obligation

### Different hypervisors

#### Microsoft virtualization technology

- parent partition: logical hard drive for running hypervisor layer, and nesting
  child partitions
- Child partitions: Logical hard drive responsible for running isolated OSes
- Virtualization stack Collections of resources including the hypervisor, which
  makes up the Microsoft hyper V
- Virtualization service provider: parent partition side, server that handles IO
- Virtualization child provider: child partition component, that sends request
  to VSP
- VMBus: send requests and data between VMs
- WMI: Windows management instrumentation, manages virtual devices
- Windows kernel: manages IO request from parent and child
- IHV: hardware drivers
- VM worker process: run and configure one VM each
- Hypervisor
    - Manages a minimum set of **hardware** components
    - **Isolate** component, to enhance **loose coupling**
    - Provide **security**
    - Enhance **performance**
- ![Image](./assets/1-2-microsoft-virtualization-technology.png)

#### Xen Virtualization Technology

- ![Image](./assets/1-2-xen.png)
- Host Hardware: real hardware
- Xen Hypervisor: lean software layer, that runs on **host hardware** and
  manages the **CPUs, memory, IO devices**
- Guest Domains / Virtual Machines
- Control Domain, or domain 0: special virtual machine with privilege to access,
  and handles IO functions and VM's interactions
- Toolstack: control stack on domain **0**, enables user to manage **virtual**
  machines

### Virtualization Technologies

- host components: real hardware
- guest components: virtual system created from host machine, includes os,
  network, databases and servers
- Hypervisor, or virtual machine manager: manages virtual machines
- Networking technologies
- Web services: manager over the internet
- Front end systems: interface between user and virtual machine

### Advantages and challenges

#### Advantages:

- Deployment:
    - Less **cost**, reduction in cost of hardware, energy, cooling , expertise
      and administration
    - Less time needed
    - Creating servers on different system is **easy**
- Maintainence:
    - **Scalability**
    - **Efficiency**
    - **Reliability**
    - Fast data disaster **recovery**
    - Backups and migrations are **easy**

#### Challenges

- **Security** of virtual machines
- Need **expertise** to manage virtual machines
- **Reliant** back up systems are needed
- Failures of **host** will affect the guest system.
- **Performance** may not be as good as using physical ones

### Future of Virtualization

- Is and will be the **dominant** component of cloud computing
- Decoupling of application will **increase**
- **Less** energy consumption by optimizing OS and VMM
- **Dominant** in IT market
- Even more **scalable**
- Better **security**

### Major classes of hardware virtualization

#### Full virtualization, Hardware virtual machine

- Complete simulation of actual hardware, so that OS and software doesn't need
  any modifications.
- AWS Support
    - HVM AMI (Amazon Machine Images), in a fully virtualized environment
    - Can run OS without modification
    - HW extension is support
    - Instance type support: all current
    - Regioin: all

#### Paravirtualization

- Allows guest applications to be executed, as if they are running on a separate
  system, but hard ware environment is not simulated
- Guest need to be modified
- AWS Support
    - PV AMI: special bootloader is used
    - Can run on host hardware without explicit support for virtualization
    - Hardware extention is not supported
    - Some instance types are supported
    - Some regions are supported

### Layering and Virtualization

#### Layering

- Definition: approach to manage system complexity
- Goals:
    - Simplify description of subsystem, to **abstract through interface**
    - Minimize interaction between subsystems
    - Enable us to independently manage each subsystem
- Layering in computers:
    - Hardware
    - Software
        - OS
        - Library
        - Applications

#### Interfaces in virtualization

- ISA (Instruction Set Architecture): Instruction Set Architecture: boundary
  between hardware and software
- ABI (Application Binary Interface): Allow application and library modules to
  access hardware, invokes **system calls**
- API (Application Program Interface): Is a set of rules or protocols that
  enables software applications to communicate with **each other** to exchange
  data, features and functionality