However, most data centers are more comfortable using traditional, block based storage for the bulk of their data services. Several recent surveys including one from VMware still put the use of a block based fibre channel SAN as the dominate shared storage topology for virtualized environments. Even the (distant) second most popular is iSCSI, also block based. For these data centers and many that will be selecting a SAN in the future, buying a separate storage system just to provide file services may not be the best choice, especially when those remaining file services amount to user home directories or storage for database dumps. Purchasing a separate NAS for these purposes not only has the acquisition cost of the NAS controller and storage, it also has an operational expense in that two separate systems would need to be managed and their data protected.

Option 2:  Adding A NAS Gateway

Adding a NAS Gateway seems like a viable alternative. Typically this means purchasing a separate device that attaches directly to the SAN to which some portion of the storage is hard provisioned. Unfortunately this brings an additional layer of complexity because the NAS Gateway has to be integrated into the SAN like any other attaching host would be. Since NAS Gateways are not based on a typical OS, the IT staff may not have the experience to successfully complete and more importantly, maintain this integration.

Even if integration can be achieved, since the storage often has to be hard allocated to the NAS Gateway, there are still essentially two separate storage platforms that need to be managed. All that’s really happening in this environment is that the SAN and NAS Gateways are sharing the same cabinet of disks. There is often little if any integration between the two products, including gateways that are available from the storage manufacturers themselves. This is evidenced by the fact all the feature sets, like snapshots, replication and provisioning, are disabled for the portion of storage that the NAS controls. Then the NAS re-provides these services. In some cases this may be acceptable, in others, the SAN platform may have provided better and more capable versions of these services. In either case the feature set is being bought twice.

This is especially ironic when SAN storage providers that have invested much of their development resources in designing systems with advanced snapshots, provisioning, replication and data movement, decide to enter into an OEM relationship with a NAS vendor. As a result, in their desperation to provide NAS services they end up having to turn off or downgrade much of the intellectual property that made them so unique as storage vendors.

The hardware NAS Gateway has the same problem that a stand alone NAS does when it needs to be integrated into a virtual server environment. What to do with it? If it is just going to provide file services then it has to be managed separately and outside of the virtual server environment. If it’s going to host the virtual environment then much of the investment in the SAN goes to waste.

Finally, and maybe the biggest single problem, is the cost of the NAS Gateway, especially when it’s provided by the manufacturer of the SAN itself, as it often is now. Assuming that the block based SAN is used to host the majority of mission critical applications and the virtual server environment, using an expensive hardware based NAS Gateway for basic file services seems like a waste of money. This leads customers to consider the third option, using a traditional server as a file server.

Option 3: File Servers

The NAS was originally designed to replace a server which was running a general-purpose OS, providing file services. The success of NAS was a direct result of this specialization, and it could typically provide the same level of performance as three or four general-purpose file servers. Despite this fact, these types of file servers remain a popular option, maybe even more popular as server virtualization has become further entrenched in the enterprise.

The first reason for this is general familiarity. Typically, the OS loaded on the file server is what’s already being used throughout the data center, and there is often limited cost to stand up a new file server. All that’s needed is the server hardware, internal or SAN attached disk and the operating system, which has often been pre-paid thanks to the corporate licensing practices prevalent in the enterprise today. And, when this new server is actually just a virtual machine, much of the hardware costs become irrelevant as well. The virtualized file server also benefits from easier implementation into the shared storage environment since it will ride on the work done to integrate the virtualization host.

The challenge with the virtualized, general-purpose file server is that, like a stand alone file server it requires more resources than a single purpose NAS, something which is a high concern in the virtual environment. Also, like the stand alone version, most environments will require more of these non-specialized servers in order to keep up with increased user counts and increased storage capacity needs. Like with application servers, the economy and simplicity of virtual file servers may actually be a weakness, as file server instances can proliferate almost unabated.

Each of the above options has its own strengths and weaknesses. The fourth option is a virtual NAS server that can integrate cloud storage as its back-end to leverage the strengths of each of these previous options, while eliminating many of the weaknesses. Companies like Nasuni with their File Virtual NAS are well down the path to providing this ‘best of both worlds’ solution.

Option 4: Virtual NAS Services

The emerging fourth option to providing file services is the Virtual NAS appliance. It couples the focus of the single-purpose-OS NAS with the virtualized flexibility of the general-purpose-OS file server. This combination allows the virtual NAS to serve many more users while requiring less disk capacity than the traditional, general-purpose file server, but because it is virtual it does not require dedicated hardware. The typical delivery model is as a software based NAS appliance that can be easily added to a virtual infrastructure and made available for serving files. Virtualization gives it the capability to migrate the NAS appliance to different physical servers as needed, or even migrate the NAS's disk image to a new storage device altogether, providing both availability and load balancing.

Another important benefit is that now the full capability of the SAN system can be leveraged, not replaced by the virtual NAS. This means the investment that the storage manufacturer made in snapshots, replication and other capabilities does not go to waste or need to be purchased twice. It also means that the virtual NAS provider does not have to waste valuable development time recreating features that it can assume will already be there.

Companies like Nasuni are taking this capability a step further by leveraging cloud storage to further optimize the virtual NAS. In this use case the on-site storage acts as a cache that holds the active data set. As data is created or changed it is copied to the cloud on an hourly basis, creating a near real-time backup of the data. While the file listing stays in local cache, the oldest local copies of data are removed so that local capacity does not need to be expanded. This is ideal for the traditional file services environment where most data is only active for 30 days and then becomes almost permanently dormant. If access to an older file is needed, that data is recovered transparently from cloud storage, for which the user only notices a slight delay in access.

This migration - denigration capability provides two valuable functions to the users. First, the SAN storage capacity allocated to the virtual NAS can be set one time, no need to revisit the virtual NAS's data store and expand it. Setting it and forgetting it greatly reduces the administrator time associated with the storage management aspect of operating local storage.

Second, as Storage Switzerland shows in its latest test drive of the Nasuni Filer, this becomes the ultimate disaster recovery capability. Recovering the virtual NAS is a matter of reinstalling the software, which takes only minutes, and the security keys that access the cloud version of the dataset. At that point the filer is back up and running with all the data available but still in the cloud. That data is then recovered from the cloud as it’s accessed instead of having to wait through a long recovery process.

Virtual NAS services, especially with a cloud back-end, provide a complete file services solution for the virtualized environment that’s fully protected, yet leverages the capabilities of the SAN investment. Because it is designed from the ground up to work in a virtual environment, it is resource friendly and doesn’t add features that the virtual environment already has, like high availability. A virtual NAS should also bring down the cost of managing file services since it becomes part of the virtual environment, not a separate island outside of it. Finally, back-ending this service with cloud storage also reduces operational overhead because the local volume will likely never need to be expanded or backed up since data is constantly being replicated to the cloud.

Nasuni is a client of Storage Switzerland

George Crump, Senior Analyst