Ceph does not deal very well with, or takes a long time to recover from, certain kinds of network failures and individual storage node failures.

I believe the community that supports Ceph is working on this. They will be providing solutions to improve these issues in the newer versions, like Jewel and in the future with technologies like BLU Store and RDMA.

Red Hat Ceph Storage is difficult to maintain. We use CLI tools for maintenance, and the concept seems challenging. Additionally, it is difficult to expand the product due to balancing errors. It takes some time to re-balance the storage in case of server failure. They could improve the speed of the process.

The operational overhead is higher compared to Azure because we own the hardware. It would be nice to have a notification feature whenever an important action is completed.

If you use for any other solution like other Kubernetes solutions, it's not very suitable. It's very good if you have a customer that is using a Red Hat deployment like OpenStack or OpenShift. It's not great if you aren't. They need to adjust the product in order to have compatibility with other solutions also.

The installation of the software could be simplified. 

What could be improved in Red Hat Ceph Storage is its user interface or GUI.

It took me a long time to get the storage drivers for the communication with Kubernetes up and running. The documentation could improve it is lacking information. I'm not sure if this is a Ceph problem or if Ceph should address this, but it was something I ran into. Additionally, there is a performance issue I am having that I am looking into, but overall I am satisfied with the performance.

Ceph Storage lacks RDMA support for inter-OSD communication. That is a huge loss in terms of performance. It's also very intensive on the backend network and takes a lot of resources from the network. I'd like to see a higher performing CephFS, which is the file system part of Ceph.

We have encountered slight integration issues. 

This product uses a lot of CPU and network bandwidth. It needs some deduplication features and to use delta for rebalancing. 

After adding new OSD, e-balancing is an huge issue if not properly configured, as it can cause huge performance issues.

The storage capacity of the solution can be improved. 

Rebalancing and recovery are a bit slow.

Ceph is not a mature product at this time. Guides are misleading and incomplete. You will meet all kind of bugs and errors trying to install the system for the first time. It requires very experienced personnel to support and keep the system in working condition, and install all necessary packets.

It needs a better UI for easier installation and management.

Routing around slow hardware.

An area for improvement would be that it's pretty difficult to manage synchronous replication over multiple regions. I also don't like the containerization method for Cephadm in the recent deployment. In the next release, I'd like to see reports for security and performance.

Ceph lacks a little bit only in performance. It needs to scale a lot and needs very fast and well-orchestrated/configured hardware for best performance. This not a downside though, it is a challenge. Ceph only improves the given hardware.

In the deployment step, we need to create some config files to add Ceph functions in OpenStack modules (Nova, Cinder, Glance). It would be useful to have a tool that validates the format of the data in those files, before generating a deploy with failures.

I would like to see better performance and stability when Ceph is in recovery.

Geo-replication needs improvement. It is a new feature, and not well supported yet. 

The management features are pretty good, but they still have room for improvement.

The solution needs to offer support for the fiber channel as a protocol.

Please create a failback solution for OpenStack replication and maybe QoS to allow guaranteed IOPS.

Some documentation is very hard to find. The documentation must be quickly available.