Storage Switzerland Drives SME Scale Out NAS
Recently Storage Switzerland began testing the Gridstore in our labs. The Gridstore is a scale out storage system focused on the small to medium sized enterprise (SME). Scale out storage has brought a lot of advantages to the large enterprise data center in the form of near limitless scale of performance and capacity.
The scale out concept would be an ideal fit for the SME data center as well but some of the complexity and cost has to be driven out of the enterprise scale out equation. Gridstore is attempting to do just that; bring enterprise class features to the SME data center but make them affordable and manageable.
Thursday, August 18, 2011
With that in mind we brought the Gridstore into our labs. As is the case with other products we test, this is not a short term ‘see how many things we can learn in a few days’ test. We are going to live with the Gridstore units for at least 90 days. Stay tuned to Storage Switzerland for continued updates on our use of the system. The first step of course is to get the system up and running.
Initial Setup
Like most scale out storage solutions the Gridstore 2.0 NAS is comprised of servers, called nodes, that when connected together are aggregated to create a grid of storage. You don’t manage an individual node though, you manage a single entity that is the grid. Since these devices are targeted at the SME, space and power efficiency is critical and the Gridstore NAS meets the demand well. We were stunned at how quickly we were able to setup, configure and start writing data to these systems.
George Crump, Senior Analyst
Test Drive
Gridstore is a client of Storage Switzerland
An important point when planning on a Gridstore or any scale out storage system is that they do consume ports. The good news is 1GbE ports are relatively inexpensive.
As you can see in the video below, after the units were installed and powered up, it was time to load the management software and begin configuring the systems. The management software comes as part of a MMC plug-in so integration with Windows is very tight. As I stated in the opening Gridstore is targeting the SME space and the busy SME IT administrator who does not have the time to chase down IP addresses and manually add and configure nodes. Fortunately with the Gridstore software they don’t have to. In our tests (we did it twice) the nodes came up in DHCP mode and were auto assigned an IP address. The software automatically found the nodes and asked if we wanted to add them to the storage pool. We also had the option at that point to set the current IP’s to a static number and to configure email alerting.
Having been involved in several scale out implementations, this is the easiest that I have seen thus far. In fact the only added complexity was the time it took to cable up the nodes, which was maybe an extra five minutes. We had the systems from box to accepting data in less than 30 minutes.
With the systems configured, we proceeded to follow the wizard to create a volume and assign access to that volume. Our other Windows clients were then provided access by installing the Gridstore virtual controller software on them.
There may be some concern about installing software to access a NAS. Configuring the environment this way allows for better performance utilization of all of the available resources. We have not seen any ill effect on performance or stability by the use of the Gridstore software.
We count the use of the Virtual Controller software as a net gain. To date, most scale out storage has been cluster based, which requires high-performance storage nodes and complex configuration of clustered file systems, failover, backplane networks, etc. Gridstore significantly reduces the complexity and cost of clusters through the Virtual Storage Controller software that is distributed to every client machine that accesses the grid or what Gridstore calls the “Virtual Storage Pool”. This is done using parallel I/O directly to each storage node to increase throughput. What makes Gridstore unique is this combination of a grid architecture with virtualization to eliminate the need for clusters. This allows them to take advantage of parallel processing and I/O which should significantly improve throughput.
Storage Swiss Take
As stated earlier total time to set the system up was less than 30 minutes from box to writing data. Initial data writes are showing excellent performance and we are unable to impact network performance to the units. The aggregated performance of multiple nodes is going to be challenging for us to push to the limits but we are going to try. I suspect that the average mid-range data center won’t come close to maxing out the performance of these units and even if they did all they would have to do is add additional units..
In our next test we will take a detailed tour through the management interface and show some initial read/write tests.
Initial Set Up of Gridstore 2.0
As you can see from the attached picture we were able to squeeze all the units into a single shelf in the lab. The ATOM based nodes are whisper quiet and draw very little power. With the units mounted in the lab we connected each node into our NetGear 24 port managed switch and had no issues accessing the nodes. Using a high quality switch like NetGear’s is critical for this type of system.
