• In what situations would I deploy parallel NAS for a customer?
• Are there different kinds of parallel NAS?
• What are some ways to address the scale-out requirement?
• How will my customer's users experience parallel NAS?
• Do I need to get specialized parallel NAS solutions for my customer's storage, or can I reuse some legacy NAS?
• How long would a typical deployment take?
• My customer is curious about being "green," for both environmental and cost-efficiency concerns. How green is parallel NAS?
• What other gotchas should one watch for in the context of selecting a parallel NAS solution for a customer?
In what situations would I deploy parallel NAS for a customer?
You would typically deploy parallel NAS in only two situations. One is where there is a high NAS performance requirement, beyond what can be provided by a single appliance. Let's call this high-performance NAS. The second situation, scale-out NAS, is where NAS capacity requirements are growing rapidly and scaling out. If requirements are in the 30 TB range, there are plenty of effective NAS appliances you can sell that easily support 12 TB to 50 TB in-the-box capacity.
Are there different kinds of parallel NAS?
The first generation of parallel or clustered NAS came from companies like BlueArc and Spinnaker, which is now a subsidiary of NetApp. This high-performance NAS delivered file storage suitable for demanding applications in research, seismic analysis, etc. The buyers were typically scientists and highly qualified engineers, and the resellers were specialists who could "tune a Ferrari."
The second generation of parallel NAS addresses the scale-out capacity requirement. This requirement has been growing in the past three to four years, and unstructured data or file data growth is proceeding at unprecedented rates as information of all types gets digitized. While NAS-per-terabyte prices have been dropping, a very real challenge is the ability to administer all this NAS storage without requiring an enormous IT team. Storage administrators have been heard to say, "I love my first NAS appliance; I hate my fifth," and "I seem to spend a lot of time trying to figure out where my files are, which application to point to which NAS box and why one NAS box is under extreme load stress whereas all the rest seem to be idling."
What are some ways to address the scale-out requirement?
There are two flavors of solutions targeting this scale-out requirement.
There are NAS gateway solutions from companies like F5/Acopia, EMC/Infinity, OnStor and RelData that present a single NAS image for a bunch of NAS appliances that sit behind the scene. These certainly can help alleviate the NAS pain and can provide several Tier 1 features (such as snapshots and other advanced data protection techniques and many enterprise protocols and access control features). These solutions are attractive for moderate scale-out; think of eight to 10 NAS boxes fronted by the gateway. The challenge is that while these gateways provide many Tier 1 storage features and can work in heterogeneous environments, they are expensive and their performance can be constrained by the fact that the gateway itself can be a bottleneck to the appliances behind it.
The new entrant in the scale-out NAS space is cloud storage. The focus here is not Tier 1 transactional file data (for example, active engineering files) but rather digital content and archival data. This is the fastest-growing part of the storage market, and 2009 seems likely to be the breakout year. While Amazon's S3 service has made public cloud storage services, where storage is rented by the gigabyte per month, early 2009 is going to see this technology available in packaged form that can be deployed at enterprises as private clouds and to solution providers looking to compete with Amazon S3 as public clouds. EMC's Atmos is a hardware/software solution, whereas ParaScale has announced a software-only model where the reseller can package up a full solution with Linux hardware of their choice. ParaScale's channel-focused model is noteworthy for its margin: Resellers can gain margin on the base software and then tack on margin on hardware. There are also other startups in this space that provide "cloud-like" storage, but you have to look closely to ensure that the particular NAS protocols of interest scale out well with capacity demands.
How will my customer's users experience parallel NAS?
The whole point is for the NAS cluster to appear like a single NAS appliance -- the user does not see the component parts. The technical term is "single system image." From a user perspective, the better solutions will be completely transparent to the user. In the real world, customers have to rely on a systems integrator or VAR to see what vendors support particular access methods (for example, providing Windows and Linux users access to the same file data).
Do I need to get specialized parallel NAS solutions for my customer's storage, or can I reuse some legacy NAS?
If your customers are looking for parallel NAS for the highest performance needs, you really need to design for performance from first principles. You should feature a product mix from quality vendors with widely adopted methods rather than more avant garde approaches. Gateway NAS and cloud storage solutions can certainly provide appealing economics for customers by leveraging some of their legacy storage, but you'll need to be trained to deliver the goods. Some of the gateway solutions are expressly designed to fold in legacy NAS solutions and can do much with the help of a knowledgeable reseller. Cloud storage solutions, the software-only kind, are also designed to leverage legacy disk, but, again, the customer is going to benefit from having you help construct a hybrid solution with lower-cost, nonbiased disk storage feeding the cloud.
How long would a typical deployment take?
It really depends. High-performance NAS, as mentioned before, typically is like "tuning a Ferrari." However, some vendors, such as Isilon, have done a good job making the process relatively painless. Gateway NAS solutions can come up fairly easily for basic functionality, but these are feature-rich, and the final configuration can take some time. Cloud solutions, by design, have been built to be extremely simple to put together and scale to petabytes. Early indications are that solutions like ParaScale's are essentially self-configuring and self-managing for the most part by a general IT administrator. We will know more in the next weeks and months.
My customer is curious about being "green," for both environmental and cost-efficiency concerns. How green is parallel NAS?
The current generation of solutions is smart about energy and cooling, because there are performance consequences from not being smart about it. Some of the gateway solutions can work with spin-down arrays on the back end, but this varies by vendor and solution. Among the newest group, the cloud storage vendors certainly talk about spin-down and tiering, and we will have to watch this space as the solutions come to market in early 2009. Architecturally, the way true cloud storage solutions deal with metadata is particularly friendly to spinning down huge portions of the cloud. Also, software-only solutions like ParaScale's will be able to leverage the newest Linux hardware with the smartest energy saving and spin-down technologies as these platforms emerge.
What other gotchas should one watch for in the context of selecting a parallel NAS solution for a customer?
While there are some NAS appliances that are very capable unified storage machines (Fibre Channel SAN, iSCSI SAN and NAS), most parallel NAS options are primarily NAS-focused. Another tricky spot is if your customer has demanding administrator access requirements (for example, multiple administrator role types), mixed-mode client access (that is, clients accessing data from Windows, Linux and Solaris all at the same time with differing authorization levels). In these situations, some of the gateway solutions are your best bet. With cloud storage, the upside is extremely appealing, and the opportunity is going to be huge in 2010, but in 2009 you have to be comfortable with emerging technologies to carry this in your line card. But then again, your customers may really appreciate your bringing the affordability of cloud storage to their attention in 2009
About the author
Jeanne Wilson owns Condor Storage, a storage-focused reseller organization offering services in the United States and Canada. She has more than two decades of experience architecting fault-tolerant data storage solutions for SMBs and enterprise clients.
This was first published in December 2008