RAID learning

RAID learning

Table of Contents

  • RAID's benefit
  • RAID's types
    • type 0
    • type 1
    • type 10
    • type 2, 3, 4
    • type 5
    • type 6
  • RAID's limitation


1 RAID's benefit

So, what the RAID is? RAID is short for Redundant Array of Independent Disks. In the storage industry, the speed, redundancy, and cost are the three factors that the market cares about.

I/O speed: hard drive's I/O performance is limited by the physical law, if we continue to use magnetic storage materials.

redundancy: the fault of a subset disks in an array can't damage the whole data stored in them.

cost: the density of data per dollar.

RAID is a common solution to face that.

2 RAID's types

2.1 type 0


 parallel multi disks together, with the highest disk write/read IO performance, but none data security.

split the data into n pieces, n equals to the array size of the disks. And both the write and read are parallel to improve the speed, but one disk's damage can destroy the whole data.

2.2 type 1


lower performance but with the highest security.

the data was written to one disk with another disk as images, so even one disk is left after the disaster, the data is still there.

2.3 type 10

the combination of type 0 and type 1

2.4 type 2, 3, 4

type2 rarely used, type 3 and 4 are abandoned without any benefit compared to type 5.

2.5 type 5


type 1 is high speed but with none redundancy, type 2 are the highest security but without speed, also the most expensive cost. So type 5 is the most trade off of them.

the redundancy method is different, its is based on byte level parity check. Data is split to (n-1) pieces, with one parity check piece. those n pieces are written to disk arrays synchronously and randomly, one piece to one disk. So, if one disk is damaged, the reverse parity check algorithm can restore it. Because of that, only one disk's fault is sustainable.

2.6 type 6


upgrade redundancy method of type 5, this time, it employs two verification check blocks with two different kinds of parity check methods. The only benefit is the redundancy performance can sustain two simultaneous disks damage.

3 RAID's limitation

In the desktop environment, RAID type 1 and 0 are the common solutions, for the server side, the trade-off solutions was type 5 and 6, this trade-off is balanced by the cost, security, and speed. But for the huge cloud, both of them are not the best solutions, one strong evidence is the industry leading enterprises like google and amazon, are developed their own storage stack instead of use RAID.


Notice:


 the above SVG images are a reference from https://en.wikipedia.org/wiki/Standard_RAID_levels.

Comments

Post a Comment

Popular posts from this blog

Bluedroid stack in android

Image
Bluedroid stack in android Table of Contents Bluetooth stack in android Bluedroid .VS. Bluez Porting guide of Bluedroid stack 1 Bluetooth stack in android From Android 4.2, google use Bluedroid stack as its default Bluetooth host stack, before android 4.2, its default Bluetooth host stack was Bluez, which is also the Linux distribution's default stack. So why google changed its Bluetooth stack? The reason is maybe that Bluez stack was designed for the desktop environment, so in order to port Bluez into android, android must solve some dependency issues, like porting Dbus, Linux desktop's default IPC, into android. So there are at least five processes to power Bluetooth before android 4.2, they are systems, Bluetooth APP, Bluez, hciattach, and Dbus. The damn thing is that there are bugs between them, so Bluetooth stack is the weakest part in the android system before android 4.2, yeah, it is the usual reason that android system corrupted. So, what's t
Read more

How to setup a NAT server?

Image
How to setup an NAT server? Table of Contents What is NAT? How to setup NAT host? a shell script from vbird's book 1 What is NAT? NAT stands for Network Address Transition, which is the basic function of gateway host. I try to hack Bluetooth tethering module in android this week, it include the Bluetooth Pan profile and the NAT setup process, which means in order to implement the Bluetooth tethering, first, two devices must be connected by Bluetooth Pan profile, and then the server part device acted as gateway host to supply the network forward service. In other words, after Bluetooth Pan profile connected, the server must to setup its NAT rules to forward the IP packages. So how to setup NAT rules? 2 How to setup NAT host? My memory is still fresh that at the start of my Linux journey, I setup my Linux desktop as an NAT host to provide Internet services for my lab's classmates. During the days as a programmer, I once did a job to hacking android&
Read more

Network programming in elisp

  Network API in elisp Table of Contents 1 Introduction 2 A example 3 Http server implemented by elisp 1 Introduction Emacs is not just an editor, Yes, that's true. Emacs lisp can do network connection programming by its network API. But there is something difference with other kinds of programming language. First, elisp do not support multi-thread, pretty cool, huh!! Yeh, Multi-threads programming is a horrible venture for you, right! Second, There is no socket similarly conception in elisp, and elisp provide network programming ability as processes. Elisp can open a network connection by function – open-network-stream .  It can make network connection server listen a special port by function - make-network-process . Both of those function return a process object. Then you can communicate with the process by process's API. 2 A example I write a snippet code to figure out How to use it. ( require ' cl ) ( defvar my-network-client-b
Read more