If you are like me, you see a lot of products come and go. Further, a lot of the different products really just present the same technology. I’ve always been a hardware geek, and when it comes to storage I find myself continually learning. One of the more evasive technologies, especially in storage, is a grid.

Grid computing has a common example in the Oracle RAC (Real Application Cluster). This architecture has many servers pooling compute capacity together for the same database environment. I’m wondering who will be the first mainstream virtualization technology to provide a grid virtualization engine — 488 core virtual machine anyone?

Grid storage is here and has been around for quite a while. The issue here with storage grids traditionally is that they are very, very large. They also can incorporate multiple geographic locales, which is nice for some design requirements. There are more options available today for grid storage, including options for the people like you and me.

I recently had a chance to get the scoop on Gridstore, who makes a scale-out storage solution that is very affordable and flexible. The core of the architecture is a low-cost storage device that has its own CPU and memory resources. There are three or more of these in the grid, and the systems that consume the storage have the virtual controller software running to present the storage to them. The storage resources pool together, and scale out very wide. Figure A below shows the architecture:

Figure A

The Gridstore architecture connects Windows systems to a scalable storage grid.

In this simple example, the two Windows systems (more operating systems are planned) connect to the storage grid. Each node on the grid has CPU and memory resources, which avoids an important pitfall of scalable storage. Many storage systems that simply “add a drive tray” to an existing controller have diminishing returns on the performance, no matter what they say.

Here’s another important note about this grid storage architecture: Gridstore doesn’t use RAID. The redundancy is delivered through the nodes themselves. Now before you freak out, Amazon cloud storage resources (such as S3) don’t use RAID either. The data footprint is distributed across nodes as the storage profile grows, and adding nodes is very easy and inexpensive. So, in this regard you can ensure performance of the environment over time without a forklift upgrade.

Every data profile has a use case, and every requirement is different. What is important, however, is to ensure that the right data footprint has the right home. Does a storage grid sound like something that would benefit you? Share your comments below.