Storage
Questions and answers regarding this topic can be found under FAQ → Storage.
For a detailed description of how to create and mount volumes using the dashboard, see
Tutorials → Managing data storage.
HeiCLOUD offers block and object storage. Each instance has an ephemeral (existing only during the lifetime
of the instance) 30 GB block storage root disk. If data needs to be saved persistently and independently of
a single instance, separate volumes or the object store must be used. Alternatively, any external data
sources (e.g. SDS@hd) can be integrated via the network.
Volumes are similar to hotplug SATA or SAS disks, which can be added or removed dynamically at runtime. A volume can range
in size from one GB to several hundred TB. Currently a single volume cannot be connected to multiple instances at the same time.
Volume types
| Name | IOPS (max)* | Throughput (max)* | |
|---|---|---|---|
| HDD-Normal | 250 | 90 MiB/s | |
| HDD-Performance | 500 | 200 MiB/s | |
| SSD-Normal | 2.000 | 250 MiB/s | |
| SSD-Performance | 10.000 | 1 GiB/s | |
| *The size of a volume has no influence on the IOPS or throughput limit. | |||
In the background all data stored in volumes is replicated three times and stored on three different systems in a Ceph based storage cluster. This offers strong protection from loss of service or data.
All data in our SSD volumes is stored exclusively on NVMe based SSDs. NVMe is a protocol used by
high performance SSDs and offers significantly increased speed compared to older standards.
These volumes are therefore ideal for read/write heavy applications like databases. They can also be used as fast root drives
for your instance's operating system.
Throughput and IOPS limits for multiple volumes per instance
If multiple volumes of the same type are mounted on the same instance, the maximum throughput and maximum IOPS
of the volume type is shared by those volumes. If the limit is reached, further read/write operations are allocated evenly to
the volumes using the round-robin method.
This means that a single instance can perform a maximum of 13.000 IOPS: 12.750 IOPS (Sum of all volume types) + 250 IOPS for the
ephemeral root drive.
Similarly the maximum possible throughput is 1564 MiB/s (Sum of all volume types) + 90 MiB/s (root drive) = 1654 MiB/s.
There is no project-wide limit for IOPS or throughput.
For example if you mount one or five SSD-Performance-type volumes on one instance, it will be limited to 10.000 IOPS and 1 GiB/s total with both configurations:
- Instance with 1x SSD-Performance = max 10.000 IOPS / 1 GiB/s
- Instance with 5x SSD-Performance = max 10.000 IOPS / 1 GiB/s with round-robin distribution over all volumes when the limit is reached.