Cloud computing theory and practice Flashcards
Summarize the content of the homonym book
5 characteristic of cloud computing
On demand self service access
Broad network access
resource pooling
rapid elasticity
measured service
Entities involved in cloud computing
customer
provider
carrier (ISP)
broker (manage use performance and delivery of cloud services)
auditor (performs security audit)
DBaaS
Database as a service, db are stored in the cloud provider infrastructure, provides scalability self service
IaC
Infrastructure as code, a way to automate cloud infrastructure through code, most used tools are ansible and terraform
S3
Simple storage system, service from AWS, stores data in objects inside buckets
EBS
Elastic Block store, provide storage volumes to EC2 instances, an instance can have more volumes but not the opposite
Lambda
AWS service that provide serverless execution of code on demand
Availability zones
Datacenter with 50 000 to 80 000 servers, each region has at least 2 AZ
WSC (Warehouse scale computer)
Computers in cloud provider’s datacentes composed by cells and racks, WSC can be managed through Borg (Cluster management system)
Borg
Is a cluster management system for WSC, it include a Borgmaster (controller) and Borglets (controlled machines)
Kubernetes
it’s a cluster manager for containers, includes a master server that manage workload on minions or workers
Kubernetes scheduler
It allows management of workers resources by the master server
Kubernetes Pods
Are groups of container used as a unit of scheduling, deployment and replication, pods allow multiple containers to run an application
2 types of VM
Process VM - executed for a specific process once terminated the VM will be stopped
System VM - VM with an OS and many processes
Hybrid hypervisor
hypervisor that shares the hardware with another os
QEMU
QEMU can virtualize or emulate any components on a physical machine
Virtualization vs emulation
1) virtualization is based on an hypervisor that abstract the hardware so that it can be shared between multiple VMs, the VMs run on a portion of the hardware
2) Emulation is the process with which you alterate the hardware via a software to make a program that didn’t support the hardware run on it
QEMU 4 modes
- usermdoe emulation for code compilation and debugging
- system emulation supports VMs on a physical emulated system
- KVM hosting supports deployment and migration of KVM images
- Xen hosting emulates the hardware and allows execution of VMs with Xen
KVM
Kernel based virtual machine, is an hypervisor buildt with linux kernel
KVM components
KVM-loadable module provides management of the hardware
KVM platform emulation provided by a version of QEMU
KVM advantages
Open source
Performance
Cross-platform
Simplicity
Paravirtualization
The guest OS knows it’s running on a virtualized environment that allows VMs to communicate with the hypervisor via API, that results in better performance
Xen
Hypervisor that use QEMU as device emulator and supports paravirtualization
Nested virtualization
situation where an hypervisor is a guest OS of another hypervisor
SAN
Storage Area network, interconnect server with storage that uses fiber channel
Fiber channel
Protocol based on 7 layers:
- Physical layer
- Transmission protocol (encode/deocde data)
- Signaling protocol (framing and flow control)
- Common services layer
- Protocol mapping layer
Fiber channel uses ID instead of a mac address
CDN (Content delivery network)
Offer better performance with replication and caching of data on servers, 2 types of servers:
- Origin servers -> data updated by the content provider
- Replica servers -> cache data and serve as data source for clients
Storage models
models that describe the layout of a data structure in a physical storage, 2 commonly used models:
- Journal
- Cell storage
Journal vs cell storage models
Cell storage divide the storage in same size cells and objects to store are the same size of cells
Journal storage consists of a manager and a cell storage where the entire history of a variable is stored instead of just the current value
DBMS (Database management system)
system that can manage the database, it supports a query language that is a dedicated programming language for interacting with databases
NoSQL
Is a DBMS that doesn’t rely only on SQL, is mostly useful with big quantity of data that don’t need a relational model
NoSQL databases type
Document store -> memorize documents in JSON, XML or BSON formats
Key-Value
Colummn-family -> stores tables with many rows and few columns
Graph -> graph datastructures used to represent and memorize data
Distributed File System (DFS)
It’s a way to group data logically in the same place even if distributed across the network. An example of DFS is NFS.
NFS (Network file system)
DFS based on UNIX, has 3 major characteristic:
- Layered design flexibility
- Hierarchical design for scalability (files in direcotries)
- Systematic design with metadata (file owners, access rights)
Stateless servers
Servers that don’t require a connection to be made but respond directly do clients, client requests are seen as independent transactions
