Pages

Sabtu, 02 Maret 2013

First look at grid storage with Gridstore


Takeaway: There is no shortage of new and different ways of housing data today. In this blog post, IT pro Rick Vanover takes a look at one of the new grid storage architectures.
Last fall, I first came into contact with grid storage. This approach of adding nodes dynamically to a storage pool with a low entry cost and low incremental costs is one that I find compelling. I recently had a chance to play with the Gridstore series of storage products, and it is a different way of doing storage.
That’s a good thing, however. As it turns out, this is a very easy-to-use storage solution that really anyone can handle. The Gridstore nodes start in a configuration of at least three nodes, and can dynamically add nodes as performance and capacity dictate without the upfront investment.
With the first three nodes, the configuration goes over multicast Ethernet for easy discovery and then specific network configuration can be set subsequently. This is all done through an MMC application in Windows, and was quite intuitive. The configuration interface is shown in Figure Abelow:

Figure A


This interface is pretty straightforward to use, and each computer that has the Gridstore software installed on it can be created a vStore, which is analogous to a volume and will inventory as locally attached storage (though it is accessed over the Ethernet network).
The data profile of this virtual storage grid is distributed across the present nodes, and nodes can fail (an important protection element!) as more nodes are added. In the 3 node example, one node can be accommodated as failed.
To see the data distributed across all of the nodes of the grid, inside the MMC console you can see performance of each node and of each node compared to each other. This can be a good way to ensure that all nodes are performing well, and they all should perform roughly equally. If not, this may be a sign of incorrect network setup. Figure B shows the performance view:

Figure B


The interface of this tool is quite easy to use and I appreciate the dynamic aspect of provisioning storage in this case. I don’t know about you, but for some data profiles, we don’t always know what we will need. One good example in this category is disk-based backups.