Community Article #7


Hi there, I’m just back from the Tet holiday, the Lunar New year in Asia and it’s nice to write something about LizardFS, which in my opinion was the best-distributed software storage when I was looking for a solution to our storage issues.
My name is Nhan Pham T. I live in Ho Chi Minh City in Vietnam, which is UTC + 07:00. I’m a dad, a blogger, and a dreamer 🙂
I speak Vietnamese and English. For coding, I use Python, Bash, and some PHP for daily tasks or small pet projects.

I started my professional career as a Python developer and have spent more than 6 years using Linux and open-source software. I don’t actually have a personal laptop or MacBook just my work computer (strange I know). I started doing research on distributed file systems when I tried to learn Docker in early 2013. I built a small lab with Ceph and used it for containers. A few years later, I started using LizardFS for our projects in the biggest Vietnamese technology company, founded in 2004, specializing in digital content and online entertainment, social networking, and e-commerce.

LizardFS was chosen as our solution because it’s flexible, efficient, and easy to set up. We have been using it since April 2019 as a distributed file system for our product. Before that, we used a physical disk for all storage with RAID 5, 6 for files, and objects. We tried GlusterFS, Ceph, and some solutions from HP and Dell in our lab, but LizardFS has URAFT for high metadata availability, something like MooseFS Pro, but with more features and functionalities available on open source.

LizardFS is easy to learn but its documentation is not up-to-date, and a lot of issues on GitHub that are open and seem not to be closed or fixed.
What I dislike about distributed file systems is that they are generally a nightmare for system engineers like me. Lots of issues and some incidents, not so much with LizardFS though.

Thinking ahead I would like to see multithreading implemented in the next release. (https://github.com/lizardfs/lizardfs/issues/476) and a webmin to administration (https://github.com/lizardfs/lizardfs/issues/546)
Also reducing memory usage of the master server would be nice. In my case, I have 256GB RAM for the cluster but I don’t think it is enough to lead many chunkservers to exabyte levels.

In the end, I would like to thank Mark Mulrainey for starting this exciting movement of sharing the community experiences with LizardFS.