Convoy GlusterFS performance

Hi,

My goal is organize shared storage of data between 3 hosts. For this, I have used Convoy GlusterFS.

I have set up all by this article (http://rancher.com/how-rancher-storage-services-unleash-the-power-of-software-defined-storage/) and tested on my project.

My project works with data (create, copy, remove, folders with sub-folders and files). I have faced with one issue - slow performance of shared storage when manipulate with data.

Any ideas, how to solve this issue?

Thanks

This is going to come down to your network equipment and speeds. If you are sharding/replicating data between servers, it’s not going to be as fast as local disk unless you have some 10bge or fiber networking.

What kind of speeds are you getting?

All 3 hosts are located on Digitalocean datacenter (Amsterdam region). Speed is 1Gbit/s between hosts.

Yea, you are on boxes that are shared with many people and the disk performance suffers.

Before disk performance the network is going to be the bottleneck. 1gbe is good, but remember those networks are busy and the box’s network card(s) are busy as well.

I wouldn’t blame GFS or Rancher for these issues, have you tried on your own hardware and done benchmarks?

Thanks,

Hello,

I am experiencing performance issues as well in terms of HIGH cpu usage and huge latency while running glusterfs+convoy to create a share volume accross hosts that run apache+php for wordpress and other not so fancy websites. I’ve read opinions that glusterfs isn’t recommended (performance and latency wise) when having to deal with small files all the time (like I have to do).

Any suggestions on how to tweak and make it faster and reduce the resource consumption? glusterfs server + nfs client maybe (as some people around the web suggest)?

Thank you!