| index | FastDFS | NFS | Centralized storage devices such as NetApp, NAS |
| Of linear expansion | high | difference | difference |
| High-performance concurrent access to files | high | difference | general |
| File access method | Proprietary API | POSIX | POSIX support |
| Hardware costs | Lower | medium | high |
| Only save a copy of the same content file | stand by | not support | not support |
Distributed storage compared to traditional SAN, NAS, advantages are as follows:
1, performance
In a distributed memory reaches a certain size that will exceed the performance of traditional SAN, NAS. A large number of disks and nodes, combined with appropriate data distribution strategy, can reach very high aggregate bandwidth. The traditional SAN, NAS, there will be a performance bottleneck, once the largest expansion of capacity, performance will not change even lower.
2, Price: traditional SAN, NAS, price is relatively high. In particular SAN network devices, optical network costs are higher. In addition, future expansion also need to increase the expansion cabinet. the cost is too high. Distributed storage requires only an IP network, several X86 server plus internal hard drive can be set up together, the initial cost is relatively low. Extended is also very convenient, plus server on the line.
3. Sustainability: traditional SAN, NAS scalability is limited, it can take up to a few hundred head disk. If you want more than one PB shared storage, distributed storage only the best choice. Do not worry about scalability issues.
Disadvantages:
1, we need strong technical capabilities and operation and maintenance capabilities, and even the ability to develop user. Traditional storage out of the box, the hardware provided by the manufacturers, there are well documented and services. And many of them are open source distributed or is there to provide support services company based on open source systems, iterative version is faster, it is possible to solve the problems themselves need.
2, data consistency problems. ORACLE RAC for this type of data consistency high demand scenarios, distributed storage performance probably weaker, because the distributed structure, data synchronization is a big problem, although it is consistent technical progress, but also not as reliable data storage device stores a conventional manner.
3, stability, distributed storage environment is very dependent on the network and bandwidth, jitter, or if a network failure occurs, it may affect the operation of a distributed storage system. For example, once an IP conflict occurs, then the whole distributed storage may be inaccessible. Usually conventional memory dedicated SAN or IP networks, stability, is more reliable.