Philip Williams
on 19 October 2023
Opposites attract. Stateful and stateless.
Storage and container management systems are almost polar opposites of each other. One deals with permanently storing, and protecting data for as long as it’s needed. The other automatically manages highly dynamic workloads, scaling resources up and down as required.
More organisations are taking a container-first approach to application deployment and management, but the underlying challenge of safely and securely storing data still remains the same. Any storage system needs to be able to protect against hardware failure and maintain the availability of an organisation’s most important asset – its data.
Data growth happens alarmingly quickly, it is estimated that more than 2,500 PetaBytes (PB) of new data is created every day. Thankfully, this is distributed across many organisations so no one has to deal with this kind of growth on their own, yet. Scale-out storage systems like Ceph, are a great fit to deal with storage growth in any size of organisation, just by adding more nodes to a cluster you are able to add not just capacity, but also more compute power – so that performance and capacity scale hand in hand.
Creating data takes time and effort too, it’s not possible to just recreate some datasets if they happened to be lost – for example, photos or videos, medical records or financial transactions. The significance of losses might be different across datasets – losing a photo is less critical than the loss of a credit to your bank account. There are also regulatory requirements to consider as well, medical data needs to be permanently preserved for as long as 100 years.
How do we bring the two together?
A container management system such as Kubernetes uses the Container Storage Interface (CSI) to integrate with external storage systems for their block and file based storage needs.
This provides two interfaces for interaction:
- A control plane that is used for storage management – specifically the creation, allocation and reclamation of volumes.
- A data plane used for high speed parallel access to data stored in the storage system.
The CSI also implements storage classes so that underlying storage types can be mapped to classes based on the performance capabilities – for certain workloads it might be important that the data is stored on the fastest SSD based disks, but for others, like an archive, the data could be stored on less expensive NL-SAS or SATA capacity oriented disks.
Ceph: a secure, scaleable and trusted storage solution
From research organisations such as the Wellcome Sanger Institute and CERN to global telecom giants like Deutsche Telekom and BT, thousands of organisations the world over have turned to Ceph to provide reliable and scalable storage for their needs.
Alongside the scaling and resiliency features of Ceph another significant advantage is that it provides interfaces for multiple storage types within a single cluster, eliminating the need for multiple storage solutions or specialised hardware, thus reducing management overheads.
Try out Ceph for Kubernetes
If you want to quickly test a robust production-ready storage solution that integrates well with Kubernetes, try MicroCeph. Microk8s is the easiest and fastest way to get Kubernetes up and running. You can experiment with the latest upstream features and toggle services on and off and seamlessly move your work from dev to production.
In the latest MicroK8s release (1.28), the rook-ceph addon was included to allow for easy integration with an external Ceph cluster. In this guide we deploy a 3 node MicroCeph cluster, deploy MicroK8s, and then integrate the two to create a powerful compute and storage cluster.
Learn more
In this whitepaper, we discuss using Ceph in a containerised environment in much more detail. You can learn about the different types of storage systems, the different access methods used to store and retrieve data, the way that data can be protected, more detail on how the two systems are interconnected, and a detailed introduction to Ceph.
