for some scenarios, such as virtual machine active image storage, or virtual machine hard disk file storage, as well as large data processing and other scenarios, object storage is stretched. The file system has outstanding performance in these areas, such as Nutanix's NDFs (Nutanix distributed Filesystem) and VMware's Vmfs (VMware Filesystem) perform well in virtual machine image storage, Google File System GFs and its open source implementation HDFs is widely used to support large data processing based on the MapReduce model and is well supported for storage of hundreds of gigabytes, terabytes, or even larger files.
As a result, the future trend of file systems is more of a dedicated filesystem than it used to be, with a previous set of filesystem for all scenarios, and some for object storage or other storage patterns.
from another perspective, where is the "sweet Zone" of modern object Storage systems: 1. Internet and similar internet-based scenarios, not just because of the rest-style HTTP interface, And because most object storage systems are designed to scale horizontally to accommodate large numbers of users with high concurrent access, 2. Massive 10 KB to GB-level object/file storage, less than 10KB of data is more suitable for the use of k/v database, and larger than 10GB file is best to split it into multiple objects in parallel to the object storage system, most object storage systems have a single object size limit. Therefore, if the application has both of these characteristics, object storage is preferred.
Some people also make further development or improvement on object storage, so that it can well support the scenarios of archive backup, MapReduce Big Data processing, and even convert the interface of object storage to file system interface; Object storage systems such as Swift also support the use of a common file system such as glusterfs as a storage backend. Why do people invest in both human and financial resources in the technology of converting these object stores and file systems to each other? What is the point of these practices? When should you use these technologies?
Copyright NOTICE: This article for Bo Master original article, without Bo Master permission not reproduced.
File system sentiment in big Data environment
