February 7, 2013

Gridstore Files Down Rip and Replace

Isaac Lopez

The traditional model of jettisoning your old system and migrating to a new one in order to scale with your data is no longer necessary, according to storage startup Gridstore. 

The company claims that its grid-based approach to software defined storage allows systems to scale as the data scales, eliminating the need for rip-and-replace migrations when a system hits the ceiling on storage capacity.   

In a recent technical overview, CEO, Kelly Murphy, claims that by virtualizing both the processors and the entire storage stack, and distributing them across a grid architecture, systems can achieve “unlimited” scalability in bandwidth, capacity, processing, and protection. According to Murphy, storage and processing capability are increased by adding new building blocks (nodes) of storage to a standard Ethernet Network. 

Murphy asserts that a chief benefit to this model is that the system becomes increasingly powerful in capacity and parallel network bandwidth as each additional storage node is added.  With every node pooled, each job is distributed across the network, lightening the load per node. As the pool expands, each node does less than what it did previously, resulting in jobs being accomplished in less time.  The nodes themselves become very simple devices, taking packets off the network and putting them on the disk.

Through Gridstore’s software defined storage model, users establish as many parallel virtualized controllers as are needed to eliminate processing bottlenecks common with a central controller.  Distributing the data across a virtualized pool of storage, the system is focused on providing an “unlimited” amount of parallel processing power that is balanced to the demands of the storage system – with the assumption being the unlimited ability to grow the storage and processing pool as your system requires.

Murphy explains that using this approach you can start with the capacity that you need today, and grow to any size you need in the future.  “Effectively, there is no endpoint on your storage system anymore – if you need to double your capacity tomorrow, now you can,” says Murphy.

So what happens if a storage node fails?  Murphy says that with the Gridstore solution comes GridProtect, – a process in which the virtual controllers write data encoded in such a way that if you lose any number of storage nodes, the data can still be put back together without loss or disruption.  Murphy claims that GridProtect goes way beyond the capabilities of RAID, protecting against any combination of network failures, node failures, disk failures, and even silent bit rot.

Theoretically, the result is a system that scales as needed as a system grows as the data scales through storage nodes, eliminating the need to migrate to a new system once the current one reaches capacity.

Related Articles:

ScaleBase, 451 Question Database Rip and Replace

Objectifying Big Data Storage

Versant Throws Magic Cube at Big Data