Ionir: let us be your Kubernetes data mover

Ionir, a software-defined storage startup, came out of stealth today with the news that it has bagged $11m in A-series funding.

The company has developed Data Teleport, a container storage platform that manages Kubernetes workloads in the cloud. This software is based on Magellan, which was developed by Reduxio, a startup that appears to have died (see our 2019 write-up of the company).

Jacob Cherian

Reduxio’s website is 404ing today and at least four execs are now working at Ionir, including CEO Jacob Cherian and co-founder and CTO Nir Peleg.

Reduxio took in $50.4m in VC funding since its inception in 2012. Ionir’s investors include Jerusalem Venture Partners (JVP) alongside C5 Capital, a UK VC, which both previously invested in Reduxio. Under the “once bitten, twice shy” rubric they must really believe in the company. Also, it does no harm that enterprise container app storage is a hot technology right now, as Storage’s Portworx acquisition last week, and NetApp’s Project Astra developments, show.

Ionir thinks storage for containers needs to be containerised itself and not provided by links to external silos. Container mobility is limited without accelerated data migration.

Ionir CEO Jacob Cherian today said that with Data Teleport, “persistent volumes can be moved or copied between Kubernetes clusters and clouds in under 40 seconds independent of the size of the volume or the amount of data involved.”

The idea is that containers are mobile across systems in the on-premises and public cloud worlds. This is fine so long as the containers are stateless and have no data associated with them. However, most enterprise apps are stateful and rely on reading and writing persistent data. When the containers move the data associated with them also move. This can be time-consuming especially when the network is slow and the data large.

DataTeleport makes a clone of the data using metadata references, and ships the clone description to the target site. That takes about 40 seconds. The migrated container can then be instantiated and access the data as if it were all present in the target system (which it is not). But desired bits of data on the source system are moved over first.

Cherian blogs: “For read operations, a combination of global deduplication, compression  and smart prefetch is used to minimise the read-on-demand and the overall amount of data transferred between the clusters.”

 “All write operations (which tend to be latency sensitive) are executed on the target, dismissing the need for invalidation on the source.”

This means that the time to access data on the target system varies between fast – meaning it’s present – and slow, meaning it’s still on the source system waiting to be migrated. This is similar to data being present or not in a cache.

Ionir said Data Teleport will support future interfaces and technologies without the need for forklift upgrades.