Article ID: 117629, created on Oct 1, 2013, last review on Aug 13, 2014

  • Applies to:
  • Virtuozzo

Synopsis

In this video you are going to discover the architecture backgrounds, main terms and components of Parallels Cloud Storage.

This video is the part of Parallels Cloud Storage course:

Video

Article

Traditional Storage Types Characteristics

Direct Attached Storage

Pros

  • Inexpensive
    • Equipment is cheap
    • Low deployment costs

Cons

  • Not highly available
  • Slow recovery in case of failure
  • A lot of idling resources

Storage Area Network

Pros

  • Highly available
  • Limited scalability

Cons

  • Equipment is expensive
  • High deployment costs

Parallels Cloud Storage Approach

Parallels Cloud Storage turns local disks to distributed highly available cloud storage.

Parallels Cloud Storage turns local disks to distributed highly available cloud storage.

Terminology

Cluster

The basic component of Parallels Cloud Storage is a cluster. The cluster is a group of physical computers connected to the same Ethernet network and providing Cloud Storage capabilities.

The cluster is a group of physical computers connected to the same Ethernet network and providing Cloud Storage capabilities.

Chunk, Chunk Server

Information in cluster is divided into fixed-size blocks called chunks.

Information in cluster is divided into fixed-size blocks called chunks.

  • Chunk has fixed size (64MB)
  • All operations on files are split onto operations on chunks
  • Server which physically stores data chunks is called Chunk Server (CS)

Replica, Replication

Every chunk may have copies on several chunk servers. These copies are called replicas.

Every chunk may have copies on several chunk servers. These copies are called replicas.

  • It is possible to set amount of replicas for a particular container or virtual machine
  • If amount of online replicas below the configured threshold Cloud Storage creates required amount of copies on available CS. The process is called replication.

Metadata, Metadata Server, Master Metadata Server

It is necessary to store information about chunks - where they are placed, how many copies are available, etc. All this information is called metadata.

It is necessary to store information about chunks - where they are placed, how many copies are available, etc. All this information is called metadata.

  • Servers which store metadata are called Metadata Servers (MDS)
  • There should be several MDSes in a cluster to ensure high availability.
  • There is one MDS in charge of keeping metadata current. This MDS is called Master MDS

Client

The server which actually runs virtual environments stored in PCS Cluster is called client.

The server which actually runs virtual environments stored in PCS Cluster is called client.

Cloud Storage is available on Client machines as:

  • Mount point
  • iSCSI target

Parallels Cloud Storage Architecture

The cluster services are communicating with each other via a single Ethernet network.

The cluster services are communicating with each other via a single Ethernet network.

Here we have a logical configuration of four chunk servers, two client servers and three metadata servers. And by saying logical we mean that we may have less than nine physical machines.

Depending on resources available you may combine those roles on a single machine. For example, you may use one machine as a client and as chunk server at the same time.

Additional Information

Cloud Storage Guide

Official product page

0dd5b9380c7d4884d77587f3eb0fa8ef 2897d76d56d2010f4e3a28f864d69223

Email subscription for changes to this article
Save as PDF