After about a month of research and development, FastCFS V2.2.0 was released, and the main improvements are as follows:
1. [fstore] Using libaio to achieve asynchronous reading, the performance of random reading is significantly improved;
2. [fstore] supports the read-ahead mechanism, and the sequential read performance is significantly improved;
3. Fixed 3 bugs:
1) [fstore] Fixed a bug introduced by V2.1.0: a key bool variable was not assigned correctly when it was run for the first time;
2) [fuseclient] Fix the consistency problem of metadata cache caused by listing directories;
3) [fauth] Fix the problem of garbled characters caused by improper format modifiers for username and poolname.
In addition, the cluster operation and maintenance tool (fcfs.sh) developed by sungness and the k8s driver (fastcfs-csi) developed by vazmin have been released simultaneously, and you are welcome to test and use them (there is a link on the FastCFS project homepage).
FastCFS V2.2.0 adopts asynchronous reading and pre-reading mechanisms, and the read performance is significantly improved. So far, the IOPS of FastCFS has surpassed Ceph: sequential writing is 6.x times that of Ceph, sequential reading is 2.x times that of Ceph, and random writing is about 2 times the Ceph. For detailed test data, please refer to the FastCFS project official website: https://gitee.com/fastdfs100/FastCFS
V2.2.0 is a milestone version of FastCFS, please support your friends!
This article is shared from the WeChat public account - FastDFS Sharing and Exchange (fastdfs100).
If there is any infringement, please contact [email protected] to delete it.
This article participates in the " OSC Yuanchuang Project ", you are welcome to join and share with us.