Block storage

The block storage service provides persistent fixed-size raw data storage which can be attached to compute instances (virtual machines). Block storage volumes can be managed using the Cinder service from OpenStack and is based on Ceph technology.

The block storage has the following key features:

  • Standard performance block storage for virtual machines

  • Flexibly resizable and suitable for a broad range of workloads

  • Can be attached as a persistent disk with the possibility to create new persistent disks using snapshots

  • If a virtual machine is terminated the persistent disk is kept and can be attached to another virtual machine

Block storage objects are managed through the OpenStack API, as described in https://pannet.atlassian.net/l/c/rgu1rn11

Block storage in Beryllium

In Beryllium, block storage is based on a combination of HDD and SSD storage nodes. The measured storage KPIs are shown in Table 1.

Description

Measuring Unit

Measurement

Read IOPS 4K

ops/sec

up to 20.000

Read throughput 1 M 

MB/sec

up to 3.000

Write IOPS 4K 

ops/sec

up to 15.000

Write throughput 1M 

MB/sec

up to 1.400

Table 1. Storage performance KPIs in Beryllium.

Block storage in Boron

Block storage in Boron is based on Non-Volatile Memory Express (NVMe) only. The “non-volatile” refers to that data persists when the server reboots, and “express” to the fact that data is transported over the PCI Express (PCIe) interface.

This Solid State Drive (SSD) type of memory is much faster than traditional SATA storage, both owing to the non-mechanical read and write operations and avoiding sending the data through a SATA controller.