Recently, IBM Spectrum Scale has added a new feather in the cap “Persistent Storage for Containers “. It clearly means that customer can build their containerized IT environments with Spectrum Scale storage systems as opposed to buying new products. And the end result is that stateful microservices such as MongoDB and PostgreSQL can use Spectrum Scale as backend persistent storage when running in containers. One can specify persistent volumes using Kubernetes or IBM Cloud Private orchestrator (ICP).
Here are some traits that just make Spectrum Scale default filesystem for “Persistent Storage for Containers”
1. Container Mobility :
Spectrum Scale being a HPC distributed file system can be mounted to every nodes in the cluster which gives an active availability of data locally to all container running nodes. So as the container moves to another node (due to node failure or some other reason), data will be always available with in no time.
2. Runs in containers for containers:
IBM “Storage Enabler for container” with spectrum scale plugin runs in container. As other services of kubernetes run in containers (e.g. controller, api server etc.. ), so even if it fails, it fails for a very short time and container orchestrator spawn a new container immediately. Upgrade/Rollback are seamless with no down time is another big advantage
3. Multi Node access :
Container Orchestrator like kubernetes has volume access mode READWRITEMANY which does mean that persistent storage can be accessed from mutltiple pods (container) and multiple nodes. Because native Spectrum Scale client provides fault-tolerant parallel access directly to your Kubernetes nodes. So multiple pod can take advantage of same data in parallel with high-speed protocol for storage access than other protocols such as NFS or SMB.
4. Application data in-place analytics :
Persistent volume for pod(container) in kubernetes is nothing but a fileset on Spectrum Scale. Now as application’s streaming data will be landed on spectrum scale’s fileset only and Spectrum Scale has hadoop connector integrated with native POSIX Spectrum Scale client so data can be analyzed in place on the same storage pool without moving to another storage hardware or storage pool. Later results can also be sent back to some other service (running in container) as a result because SEC supports existing fileset to be mounted inside the container(pod) as a persistent storage .
5. Automated tiering of hot container data to faster data pools
Most common use case where admin will always want most frequently accessed files to be placed on fastest storage pool and the cold ones on slower pools. It make sense in container world to serve all application better in terms of high performance and less latency with reasonable infrastructure cost.
6. Data migration to remote site
Spectrum Scale has multiple ways to migrate data to multiple cloud and other remote sites for data availability and resiliency. Admin can use IBM Transparent Cloud Tiering, rsync and other migration utility to migrate data to remote site or cloud
7. Data Encryption & compression
Files in Spectrum Scale can be encrypted at REST using self-encrypting device or compressed using spectrum scale user defined policies.
Release Notes: https://delivery04.dhe.ibm.com/sar/CMA/SSA/080fy/3/IBM_Storage_Enabler_for_Containers_2.0.0_RN.pdf
User Guide : https://delivery04.dhe.ibm.com/sar/CMA/SSA/080fy/3/IBM_Storage_Enabler_for_Containers_2.0.0_UG.pdf
(Views presented here are my own and not of my employer’s)