Community Article #8


Hi everyone, my name is Wessam Aly. I’m from Egypt. I speak English, French, and Arabic. I’m also pretty fluent in C, and I do some coding in Python, Perl, and Go as well.
I have personally implemented and tested quite a few distributed file systems already, namely Ceph, OpenIO SDS, Minio, GlusterFS, and obviously LizardFS.
Professionally, I have used, implemented, and tested Cohesity, Rubrik, Hedvig, and Veritas CFS.

I also have experience with other storage systems:
IBM (Flash system, Elastic Storage System, Elastic Cloud, SAN Volume Controller, etc.),
EMC (Isilon, DataDomain, Unity, Clariion, DMX, etc.),
HPE (3PAR),
Hitachi (XP)
So, the most important part is, why did I choose LizardFS?
I have heard good things about its speed and reliability at scale (all out of the box, with no special tuning).
I only did a bit of testing, but so far I have discovered that LizardFS is one of the easiest and most direct implementations of an SDS solution among all products I have experience with.
I really like a lot of things about LizardFS, low system requirements and straightforward architecture are a huge plus, so are the clients available for Windows. Geo-replication is really useful too.

There are some bad things about the project too, unfortunately. I dislike master servers and metadata servers. The release cycle could be shorter too.
I would prefer for the metadata to be distributed among chunk servers the same way data is. If a clear changelog could be added together with a near-fixed release cycle that would mean a world to me.