The FSM can be deployed in a virtualized or physical manner. When deployed in a physical machine, the physical machine also supports other storage capabilities, but the overall storage performance may be reduced due to the management of the physical machine. The FSM is generally deployed on the control node of the FC or FSP. After the FSM is deployed, you can add storage nodes and compute nodes, create storage pools, and more through the fs portal. After each node is added, the installation program of the FSA agent process needs to be sent and received in the portal. After the installation is completed, the FSM can establish certain communication capabilities with each node. In the storage node or storage, the computing node has a cluster concept, and the corresponding control uni****C needs to select at least three storage nodes to deploy, so it is also a storage node, and three or more servers are deployed with MDC services. For a simple storage node, after adding it to the storage pool, an OSD process is created on it, and each hard disk corresponds to an OSD process. For the compute node, the VBS is automatically created after the block client is installed on the portal. Usually, the VBS needs to consume one process. Generally, there is only one VBS process on one server. If you want to improve the io capability, you need to improve the single-thread processing capability of the CPU. Or increase the number of vbs.
FusionStorage Manager (FSM)
The management process provides operations such as alarms, monitoring, logs, and configuration. It is recommended to deploy the active and
standby nodes.
FusionStorage Agent (FSA)
The management agent process is deployed on each node (server) to implement communication between each node and the FSM.
MDC
The business control process implements state control of distributed clusters, as well as control data distribution rules and data reconstruction rules. When creating a control cluster, the metadata management service (zookeeper) is deployed on three, five, or seven nodes, and one MDC process is deployed on each metadata management service node to form a control cluster.
The system can create up to 96 MDCs. If the number of MDCs in the system is less than 96, the system will automatically select one storage node to deploy MDC in the created storage pool. If the number of MDCs in the system has reached 96, the system will no longer continue to deploy MDC.
VBS
The service IO process is responsible for the management of volume metadata and provides distributed cluster access point services, enabling computing resources to access distributed storage resources through VBS. A VBS process is deployed on each server to form a VBS cluster.
OSD
A business IO process that performs specific I/O operations. Multiple OSD processes are deployed on each server, and one OS corresponds to one OSD process.
According to the deployment form of each process, fs can be divided into computing storage convergence deployment and separation deployment.
In the case of a converged deployment, the node deploys FSA, OSD, VBS, (MDC), and the virtual machine can be issued on the node. In the case of separate deployment, the storage node deploys FSA, OSD, (MDC), and the compute node deploys FSA and VBS.
For the fusionstorage disk, in addition to the system disk needs to do raid1 to improve stability, there is no need to do raid, add a separate disk to the storage pool on the fusionstorage portal (3108 2208 1708raid card needs to be given to each disk) Raid0 alone can be recognized). In addition, fusionstorage provides two links: scsi and iscsi. For Huawei's virtualization platform, fusionstorage provides storage capacity through the scsi link, which eliminates the need to configure bootloaders and the like; for vmware, or physical machines, Mount the disk for it by iscsi, so you need to configure the initiator. In addition, fusionstorage also supports the use of iscsi for third-party openstack, but some advanced features may not be possible due to platform reasons.
For the security of fusionstorage, it is divided into server level and rack level. Because fusionstorage usually uses three copies of the form, the original data is done as two integrity backups, its security is not the highest, but at the same time, its disk usage rate is one-third. In addition, when server-level security is selected, the replicas will be stored on different servers, and when the rack level is selected, the replicas will exist on different racks. So it's a bit more common: server-level security supports any two servers that are powered down or corrupted, while rack-level security supports any two racks of servers that are powered down or damaged.
