GFS distributed file system cluster (theory section)

GlusterFS Overview

GlusterFS Profile

• Open source distributed file system
• By the storage server, client and NFS / Samba storage gateways.
• No metadata server

GlusterFS features

• Scalability and high performance
• High Availability
• Global unified namespace
• Elastic volume management
• Standards-based protocols

GlusterFS term

• Brick: a storage node
• Volume: Volume
• FUSE: kernel module, the client interaction module
• VFS: virtual port
• Glusterd: Services

The modular stack architecture

• The modular, stackable architecture
• By the combination of modules, responsible for realization of functions

GlusterFS operating mode

GlusterFS workflow

• Application: By client or application to access the data mount point GlusterFSync
• VFS: linux kernel request is received and processed by the VFS API
• FUSE: VFS kernel FUSE data submitted to the file system, fuse data is delivered to the file system sucked through the / dev / fuse GlusterFS client terminal device file
• GlusterFS Client; transferring data to a remote GlusterFS Server via the network, and the server is written onto the storage device

Elastic HASH algorithm

• HASH algorithm by a 32-bit integer
• It is divided into N consecutive sub-spaces, each space corresponds to a Brick
• HASH algorithm of the advantages of elasticity
  • The average data is distributed to ensure that each of the Brink
  • Lazy solved according to the metadata server, and then solve the access bottleneck and single point of failure

GlusterFS volume type

Distributed volume

• No file into blocks
• HASH value of the property saved by file extension
• Supported by the underlying file system ext3, ext4, ZFS, XFS, etc.

Distributed volume features

• Files are distributed in different servers. It does not have the redundancy
• More easily and inexpensively expanded volume size
• Single point of failure can cause data loss
• Depending on the underlying data protection lazy

Creating a distributed volume

• Creating a distributed volume name is dis-volume, documents will be distributed in accordance with HASH server1: / dir1, server2: / dir2 and server3: / dir3 in
  gluster volume create dis-volume server1:/dir1 server2:/dir2

Striped Volume

• The offset of the file into N blocks (N bands with nodes), Server node stored in each polling Brick
• When storing large files, especially performance
• Does not have the redundancy, similar Raid0

Feature

• Data is divided into smaller blocks distributed to different blocks stripe server farm
• Reduce the load distribution and smaller file access speed accelerates
• No data redundancy

Create a striped volume

• Creating a named Stripe-volume of the striped volume, the file will be stored in the block polling Server1: / dir1 and Server2: dir2 Brick in two
  gluster volume create stripe-volume stripe 2 transport tcp server1:/dir1 server2:/dir2

Volume Copy

• Save a copy of the same file or pay would copy
• Because the copy mode to save a copy, so the lower disk utilization
• A plurality of storage nodes are inconsistent, then the capacity of the lowest node taking effect as a cask according to the total capacity of the volume

Feature

• All volume servers hold a complete copy
• The number of copies of the volume can be created when customers decide
• At least two blocks or more servers servers
• With redundancy

Create a copy volumes

• Create a copy of a volume called rep-volume, two copies of the file will be stored simultaneously
  gluster volume create rep-volume replica 2 transport tcp server1:/dir1 server2:/dir2

Distributed striped volume

• Taking into account the features of distributed and striped volumes
• Mainly for large file access process
• At least a minimum of four servers

Creating a striped volume distributed

• When creating a striped volume distributed called dis-stripe and arranged distributed striped volume, volume number storage server included Brink must be a multiple of the number of tape strips (> = 2 times)
  gluster volume create rep-volume stripe 2 transport tcp server1:/dir1 server2:/dir2 server3:/dir1 server4:/dir2

Distributed Replicated Volume

• Taking into account the features of distributed volumes and copy volumes
• Case for the need for redundancy

Creating a distributed replicated volumes

• When creating a striped volume distributed called dis-rep, the replication volume distributed configuration striped volume, the storage server Brink volume contained must be a multiple of the number of tape strips (> = 2 times)
  gluster volume create rep-volume replica 2 transport tcp server1:/dir1 server2:/dir2 server3:/dir1 server4:/dir2