650) this. width = 650; "style =" border-bottom: 0px; border-left: 0px; border-top: 0px; border-right: 0px "title =" mfs Topology "border =" 0 "alt =" mfs Topology "src =" http://www.bkjia.com/uploads/allimg/131228/0356055V2-0.jpg "height =" 781 "/>
Note:
Purpose:
1. The cluster design has a VM capacity of 70 to units, occupying 2 cabinets;
2. the VM can be migrated between the host machines in the KVM cluster;
Mfs cluster architecture:
1. The mfs cluster consists of the metadata server, log server, and chunkserver;
2. the metadata server area consists of four servers and two servers constitute a drbd cluster to solve the single point storage failure of metadata and log servers. The other two servers are recommended to choose servers with high performance and reliability, each server creates two virtual machines for use as the metadata server and log server, and the corresponding two virtual machines implement HA through keepalived, Heartbeat, and so on to achieve high reliability;
3. the Chunkserver requires at least four NICs for each machine. If the machine only has two NICs on board, you must add one dual-port Nic to bind the four NICs, this improves the network bandwidth;
4. Each host in the KVM cluster acts as the mfs client, mounts the file system of the mfs cluster, and places the Virtual Machine on it;
Mfs cluster construction:
There are a lot of materials.
Optimization suggestions:
1. Use more chunkservers;
2. configure more hard disks for the chunkserver;
3. Configure and bind more NICs to the chkunkserver;
4. The virtual machine uses an ide disk instead of virtio;
5. the VM disk uses the writeback mode.
This article is from the "xiaoli110 blog". For more information, contact the author!