Structure analysis of MFS file system in distributed storage System

&http://www.aliyun.com/zixun/aggregation/37954.html ">NBSP; This chapter describes the open source storage System MFS Linux, which was developed by Poles. The MFS file system enables RAID functionality, not only to save storage costs, but also to professional storage systems and, more importantly, to achieve online scaling. What readers must understand is that MFS is a semi Distributed file system.

Introduction to MFS

MFS's official website is http://www.moosefs.org/, where you can get more detailed help. The official MFS network composition and operation principle are shown in Figure 8-1.

Figure 8-1 MFS composition and operation schematic diagram

MFS's network consists of three parts: master server, CHUNK server, and client, where master server has only one, and CHUNK server and client can have multiple. The working mechanisms of the MFS read process and write process are shown in figures 8-2 and 8-3 respectively.

Figure 8-2 The working mechanism of MFS reading process

Figure 8-3 The working mechanism of MFS writing process

As you can see in Figure 8-3, the internal workings of MFS when writing data is clear.

MFS File System Architecture

MFS file system structure see figure 8-1. The entire file system consists of 4 roles:

Port Management Servers--master Server
Port Meta data Log Server--metalogger
Port data Storage Servers--chunk server
Port Client--client

The role of 4 roles is as follows:

Port Management Server, sometimes called metadata server, responsible for managing each data storage server, scheduling file read
Write, reclaim file space, and restore multiple-node copies.

The port Metadata log server is responsible for backing up the change log file for the Management Server, and the file type is Changelog_ml.*.mfs, so that it can be replaced when a Management Server problem occurs. The meta-data log server is a new service in the MFSL.6 version, which keeps the metadata log on the Management server and can be stored separately on a single server. To ensure data security and reliability, a separate server is recommended for storing metadata logs. It should be noted that the metadata log daemon is on the same server as the Management Server, backing up the metadata log server as its client, and obtaining log files from the Management Server for backup.

The port data storage server is the server that really stores user data. When you store files, you first divide the files into blocks, and then copy the blocks between the data storage servers. At the same time, the data storage server is also responsible for connecting the Management Server, follow the Management Server scheduling, and provide customers with data transmission. The data storage server can have multiple, and the more the number, the higher the reliability, the greater the MFS available disk space.

The port client uses the fuse kernel interface to hook up the managed data storage server on the remote Management Server, making the shared file system and the local Linux file system look the same.