Shared storage Technology

Shared storage Technology

n what is shared storage

Shared storage, which refers to multiple servers accessing (writing) the same partition of the same storage device. In layman's terms, the system allows multiple users (processes) to share the same file. If this function is not provided, it means that any user who needs the file must have a copy of the file, which clearly creates a waste of space.

Why do I need a shared storage system?

Building a multi-host shared storage system to enable each host to share data and transmit data efficiently, and to achieve data conservation and protection, is now a number of radio and television, Digital asset management, oil and gas, high-performance computing network systems and large-scale Web site system applications, etc. such as in the radio and television industry, the non-linear editing system, the staff of raw film recording and digital processing at the same time, you can let another staff editor and validation of film length, so that can effectively improve production speed, greatly reduce production costs.

How to realize the application of n shared storage System

At present, multiple hosts to the same storage unit for editing, mainly with a dedicated parallel file system to achieve, the mainstream has a semi-centralized file system and a full distributed file system.

n semi-centralized file system

The idea of a semi-centralized file system is that the client and MDC metadata managers (the mapping of storage space allocation, recycling, and file blocks to storage blocks) are directly connected to the SAN and can directly access the storage blocks in the SAN.

Before a client accesses a file on a SAN, it sends a request to the MDC server via the LAN, the MDC resolves the file pathname for it, completes the file fast to the storage block, and then returns the actual physical address of the storage block to the client server, where the client can access the storage device directly through the SAN. Read or write high-speed data from storage devices connected to a SAN (such as the Wave-easy end-to-end 4Gb Fibre Storage System AS800) (as shown in the following illustration).

In the semi centralized San file system, the control flow and the data stream are separated, only the control flow passes through the MDC Management Server. In this implementation, because the control flow data is very small compared to the data flow, MDC Management Server's pressure greatly reduced, has basically not become the bottleneck of the system.

n Full Distributed File system

Considering that with the increase of the client, the MDC server may form the bottleneck of the system, so the other way of implementation is the whole distributed file system should be born.

The full Distributed file system is no longer dedicated to the implementation of the MDC metadata manager, so that the system's client servers themselves to negotiate the storage (the first end-to-end 4Gb optical fiber Storage System AS1000) file use. The load from the LAN is shared by multiple client servers, so the server does not constitute a bottleneck for the system (as shown in the following illustration).

Any client panic will not cause the system to run, and the automatic rotation mechanism within the system automatically redirects the LAN requests on the failed server to other working client servers without manual intervention. The client server negotiates with each other over the LAN before accessing the storage devices on the SAN. After gaining access, the data blocks are accessed directly through the SAN, each client server can see all the storage resources, each client server can see the complete file system, and the file system through a rich lock mechanism to ensure the consistency of data sharing on heterogeneous platforms.

N the product introduction of shared storage

To ensure that a large number of client hosts share access and administrative rights to the same storage device, you need to install specialized storage-sharing management software or a dedicated shared file system. Software or file systems like IBM Sanergy, Rorke Data Imagesan, Quntum StorNext, IBM GPFS, SGI CXFS, etc., are developed for storage device sharing.

Among them, Sanergy, StorNext, GPFS, Gxfs are based on the semi centralized San file system design, Imagesan based on the whole distributed file system design.