Medium sized organizations often go through unpredictable periods of growth and designing a data center to respond to that kind of business expansion is difficult, especially when it comes to storage. These businesses are typically more dependent on unstructured data and were created when personal IT was the norm and the communication of ideas was all done digitally, via office productivity types of applications. As the business grows from a small entity to a larger one the need to share and collaborate becomes more critical and a NAS is often the first major IT purchase after the network itself.


The problem with stand alone NAS is that as the business grows the capacity and performance capabilities of that first system are exceeded and another one is needed. Now the business has a problem. Data is now scattered across two or more NAS systems and finding it slows down the collaboration process while it creates a major expense burden on the IT budget.


Scale out storage was designed to address this exact challenge. Scale out storage provides file services, but across multiple NAS appliances that are clustered together to act as one large NAS. When additional storage capacity or additional performance is needed another appliance, or “node” is added to the system and its capacity and performance are joined with the other nodes. The administrator and users see a single NAS environment instead of multiple nodes. This makes collaboration easy because users interact with one large, ever expanding file share.


The other advantage to scale out storage is that it increases data availability. Depending on the configuration entire nodes can fail and users can still access data. As stated earlier in an environment where the work product is purely digital, this can be critical. So, in the mid-range enterprise reliability may be even more critical. For example Storage Switzerland creates notes digitally, records phone calls digitally, records videos of webinars digitally, creates its content digitally and moves it to production digitally. Their is no hard copy to fall back on. This is the case in many medium sized businesses.


While scale out NAS has a lot of appeal to the medium sized business the typical scale out storage system is out of place in these companies’ data centers. Most scale out systems were designed for the enterprise which must deal with a wide variety of workloads other than just file sharing. The enterprise NAS systems they use now often support environments like server virtualization and databases. By comparison, all the medium sized business wants is to share files, with excellent performance and not have to keep adding new NAS devices as their capacity demands increase.


As a result of this mixed workload, nodes in enterprise scale out storage systems are often multi-core Intel servers with internal disks. A low end enterprise node can cost over $30,000 and most enterprise scale out storage systems need 3-4 of these nodes just to get started. While it may eventually use the capacity of this initial implementation the medium sized business is often made to pay for more bandwidth and processing performance than it will likely ever use. And, as capacity demands increase this excess becomes worse. Part of the reason is that the nodes in an enterprise scale out NAS have a lot of work to do other than just store data. They also have to account for the aforementioned mixed workloads, maintain cluster constancy and handle data distribution.


A scale out NAS designed specifically for the mid-range data center would have to develop a better way to provide performance and scalability while keeping costs down.  One way to accomplish this is to abstract the controller from the scale out storage system as Gridstore does with their NASg system. This enables them to remove the high cost Intel processors commonly used in scale out nodes. With this system the controller function is off-loaded to the clients attaching to the storage system, which, thanks to today’s processing capabilities, have plenty of CPU resources available to them. Handling their individual storage workload puts virtually no strain on the client while dramatically reducing the processing requirements and cost of the Gridstore NAS nodes.


Enterprise scale out systems don’t match the real world environment of the medium sized business. Mid-sized businesses don’t often have the high speed network to take advantage of the performance that many enterprise scale out storage systems charge for. Also, as stated earlier, the mid-range NAS workload tends to be very focused around collaboration of office productivity files, not on hosting databases or virtual environments.


Scale out storage systems that are focused on the mid-range like those offered by Gridstore do just that. They provide nearly infinite capacity and excellent data availability in a cost effective system, while distributing the processing load across the connected systems. As a result, these systems can be offered at price points that the mid-range data center can afford. They’re also able to provide scale out systems in power efficient, space efficient packages.


Architecturally these systems look similar to enterprise scale out systems but, because much of the processing load has been distributed to the clients, they can use less powerful processors that are also less expensive and more efficient. Gridstore for example uses the Intel ATOM processor (like the one found in most NetBooks). That power efficiency also enables greater density since the systems don’t need the airflow of bigger, higher temperature processors.


Make no mistake, medium sized businesses need performance and the NAS can’t get in the way of their collaborative effort. If the nodes in the storage system are going to use lower power, more efficient CPUs then something else has to pick up the processing slack. An ideal candidate is to leverage the processing power of the users’ laptops and desktops where that data is originated (written) and/or consumed (read). Essentially what has happened is the storage controller function has now been distributed across the small business network and is being handled by the client that’s actually reading and writing the data. The result is that the storage nodes have significantly less work to do other than to store and fetch data, since the clients are now part of the storage system. All this leads to the Intel ATOM processor which is ideal for them.


When scale out storage is designed specifically for the mid-range, the environment of the mid-range needs to be factored into the design process. That means leveraging the resources of clients that are already there which opens up the possibility of using a more cost effective, space efficient node. The result should be a scale out solution that really solves the customer’s problem, keeping up with capacity demands, without consuming the IT budget, nor all the space in the data center.

Gridstore is a client of Storage Switzerland

George Crump, Senior Analyst