There are a number of boxes and software solutions available to make a NAS solution that’s cloud-based. These include gateways like Barracuda Networks, Riverbed Whitewater, Ctera, Twinstrata, Nasuni, Microsoft-StorSimple, and many others. Software providers include Amazon's Storage Gateway, Ctera (which offers both), Riverbed (again), Panzura, and more.
All of these products use a cloud-based storage service such as S3 to hold the data. The hardware solutions also offer a local disk storage option for caching or even for holding a full replica of the data, which speeds up access.
Why move your NAS servers to the cloud?
First, a cloud NAS solution provides good data integrity by having the data replicated across several nodes at the cloud service provider. The default service for object storage typically has three replicas, and the third one is usually remote from the other two, giving disaster protection. With this, most NAS operations are well protected, though I would recommend periodic backup to a service like Amazon's Glacier, which is offline. The backup, like tape, protects against software or human error.
Second, being in the cloud, it is possible to access the data easily from any location. Rigging up software-based access to files from home is easy, so that remote workers or even the CEO can get weekend access to data. Implicit in this is that the data is host encrypted, and that passwords are properly managed.
Many of us have had to pass files to vendors. These might be specs or CAD files, or just invoices and quality reports. A cloud NAS enables that, since spaces can be set up that are read-only, and this gets around file size limits and having to set up ftp accesses. I haven’t found the capability for write-but-not-modify access and write but not read, which strike me as modes that might be used to control vendors' right to alter files after they’ve been sent to you, but I expect that is a matter of time.
NAS systems are big repositories, and much of the data goes stale and is never accessed again after just a few days. This data tends to clog up the file system, and makes finding current data difficult to identify. The gateway boxes move older or unused data out to the cloud, but that doesn’t resolve the data quality issue. I’m expecting extended metadata to help this, as the object stores get more tagging capability, but this doesn’t obviate the need for a set of data policies.
Moving the data to the cloud provides a good excuse for a purge, especially as it becomes possible to bill back the storage costs to each department. Since no one will be happy about dumping files, offer to keep any valuable files on Glacier or a similar archiving service, and this should save costs.
Speaking of cost, what does it cost to store data in the cloud? Today’s pricing for S3 is $95/month for the first terabyte, and Glacier is $10/terabyte/month for any amount of capacity. Since equivalent capability requires three fairly decent servers, it could take as much as eight or 10 years of S3 to match the capex and opex of those servers.
There’ll be cost for the gateway box, or for software, to offset the savings, but keeping the cloud NAS clean and moving low-usage data off to cloud archiving would keep the costs down even lower, and offset the extra purchases.
All in all, it looks as if cloud NAS is a potentially good financial choice, offers some new ways to share data, and resolves those pesky backup issues. I suspect that it will be a well-used function of the cloud, and that Amazon Web Services may join the crowd of vendors that are beginning to offer a cloud or web storage facility.
