Linux Ceph HA vServer / VPS

Our Ceph HA vServers are virtual servers that are operated on a highly available cluster with Ceph-distributed storage. This keeps your server online and your data protected and replicated even in the event of hardware failures.

Ceph is a distributed open source storage system that automatically distributes and replicates data across multiple physical hosts. IP-Projects always use at least three Ceph nodes. Ceph distributes the data between all server systems according to a hash procedure. In contrast to RAID, Ceph is also suitable for very large data volumes, distributes them across several physical systems and is flexibly scalable. It ensures accordingly:

• Redundancy of data
• Self-healing in the event of failures
• Scalability without downtime
• Load distribution

HA stands for High Availability. For our Ceph HA vServer this means:

• Automatic failover of host or hard disk problems
• Live migration to other nodes in the event of hardware defects
• Minimal downtime thanks to clustering
• Maintenance tolerance by moving the vServers to other cluster nodes during live operation

We rely exclusively on high-performance AMD EYPC systems for our Ceph HA vServer products. We also use Datacenter NVMe SSDs and a 200 Gigabit/s LACP network:

• Ceph Cluster with (at least 3) storage nodes (OSDs)
• Enterprise-Hardware NVMe SSDs
• Redundant network through LACP channel via different network switches

In general, our Ceph HA vServer products are always suitable when high data security and reliability are required:

• Need for the best possible uptime and data integrity
• Critical applications or web stores
• Databases, mail servers, container workloads
• Productive customer projects

As with most of our vServers, the Ceph HA vServer offers include daily data backups at one of our other locations. These automatic backups can be extended to a rotation cycle of up to 7 days for a fee. There is also the option of booking an additional backup drive, which can be integrated as SMBv3 within the Ceph virtual machine. In order to always achieve maximum availability, the status of the Ceph cluster is of course also permanently monitored. A defective OSD within the Ceph cluster is detected by us at an early stage and replaced by our data center technicians without a maintenance window during operation.

Our Ceph HA vServers are located on a Ceph cluster consisting of at least three hypervisors, which are shared with other customers. The KVM hypervisor of the host system handles the demarcation between the customer servers. The resources are limited per vServer, which means that it is not possible for a single customer to use all system resources. The limitation is based in particular on the IOPS (input/output requests per second) of the NVMe SSDs and the maximum write and read speed. In general, we also offer our customers exclusive Ceph HA clusters. These are then set up individually according to customer requirements and provided exclusively as dedicated servers. If required, please contact us!

As part of our Ceph HA vServer packages, it is possible to expand the virtual machine at any time. If you have reached the maximum expansion level 'Ceph HA vServer XL', our experts will be happy to provide you with an individual offer and can then support you in exporting the vServer - for example to a dedicated host environment.

Ceph stands out from other cluster file systems (such as GlusterFS, MooseFS, Lustre, DRBD, ZFS, etc.) in many ways. Through its architecture, flexibility and especially through its focus on distributed, fault-tolerant and scalable storage. IP-Projects has been working with Ceph in conjunction with high-availability cluster environments for many years and its fault tolerance in particular has convinced us. So far, our customers have never experienced any data loss even in the event of major error scenarios. In addition, simple scaling across the board is possible at any time without risk. Ceph is therefore particularly suitable for:

• Virtualization (e.g. Proxmox, KVM, OpenStack)
• Cloud-native environments (Kubernetes, Object Storage)
• High availability / Enterprise Hosting
• Fast, secure storage of large amounts of data
• Scalable infrastructures with many clients or VMs

We use Ceph as hybrid storage. This means that virtual machines can also be located on the storage nodes at the same time. There are numerous advantages to this approach:

Efficiency

As no additional hardware is required for a separate, pure storage cluster. This approach not only saves costs for necessary licenses and hardware, but also energy.

Low latency & better performance

As the VMs are located directly next to the local Ceph OSDs, the access time to the data is reduced and the network is noticeably relieved.

Resilience

Even in the hybrid use case, fail-safety remains intact. In our setup, we also ensure that a certain total size is not overwritten. By distributing all VMs across multiple clusters, you get the best performance while minimizing the risk of failure.

All Ceph cluster nodes are currently located at our highly certified server location FRA1-NTT. There are currently no plans for site redundancy in several regions, as there have been no customer requests for this to date. Technically, however, this would be possible - thanks to the large number of our data center locations - and would be the next development step in our Ceph HA vServer product line.