The original title of this article was “NAS is dead” but I thought that it was too strong, and it wasn’t my idea to restart some polemics about an article posted some weeks ago. In any case, everyone working in IT knows that the concept of death has many meanings (e.g.: also mainframes and tapes have been declared dead many times!). BTW, the goal here is to talk about a very touchy topic: internal dropbox-like services for enterprises.
Last week I spent a couple of days with a primary storage vendor talking about object storage. We talked about many aspects about it but, above all, there was a topic that is very sensitive to enterprises now… I can resume it as follows:
Damn iPads!
Every IT manager on this earth is dealing with some sort of mobile device. If not, they will soon!
Moreover, the BYOD phenomenon is making it worse. Tablets are immature from an enterprise point of view and they are difficult to manage, especially when we talk about stored/accessed data. Yes, of course there are some mobile management platforms out there but they solve only part of the problem and they don’t come cheap!
Damn Dropbox!
Obviously the problem is not Dopbox itself but consumer sync&share services in general. The perfect storage for mobile devices is the cloud storage and, especially when the device is a personal one, you tend to use the same services for storing personal and business data!
The process is simple: you pick up a file from your PC, then you attach it to an email and, finally you send it to your personal email address. The last step is an easy upload to your Dropbox-like account. As easy as diving in a swimming pool, that unfortunately is full of piranhas!
In fact, every time that someone does this operation the enterprise loses control of its files: no one knows who is in control, what he is doing with it, if the version is correct and eventually if the people who are accessing that file have the right to do it.
And I won’t talk about the level of security that you can find on a consumer grade sync&share service is probably far from your company policies.
This happens more and more in every enterprise and may of them are looking for viable solutions.
Damn NAS!
Tablets don’t show a filesystem at the user level: they don’t know about the existence of CIFS or NFS. NAS (Network Attached Storage) was born during the PC era and it has seen its time. Enterprise tablets are growing more than PCs and these devices need to access company documents: you need a different and innovative solution “to rule them all”.
The solution could be a next generation “unified storage” with the ability to deal at the same time with PCs and tablets: in practice, instead of thinking about blocks+files, it’s time to think about blocks+files+cloud… or at least files+cloud. In this case cloud means objects.
A damn simple solution!
A private “enterprise dropbox” is already suitable for all enterprises. but if it isn’t integrated with Active Directory and it doesn’t share the same content shared to PCs it’s not a viable solution.
Actually, some vendors are working on this kind of solutions, some more actively than others. End-to-end integrated products and solutions are around the corner, and in all the solutions that I’ve already seen in the last two years the convergence is clear. Object storage is the backend (both in private and public products/services) while, on the front-end, we find different gateways allowing a more and more seamless experience between PCs and tablets. In the end, the user obtains an experience that meets his expectations while the company efficiently maintains a better control on its data.
Bottom line
NAS, as we know it, doesn’t solve our sharing needs anymore as it did in the last decade: it’s necessary to find new and broader solutions. New mobile clients are side by side and soon they’ll outpace traditional PCs. Majority of Enterprise are dealing with this kind of revolution, and if they haven’t yet they will soon. It’s only a matter of time: bye bye, ordinary NAS!
Enrico NAS is not dead per say, in fact many object storage based products and services are accessed via NAS (NFS, CIFS or other file) based methods, in addition to S3 and other REST APIs, and on an increasing basis, IOS or other APIs/personalities. Likewise, deep down hidden underneath many of the object storage and cloud storage software stacks still exist filesystems such as btrfs among others that help to abstract the underlying block storage from the object storage abstractoin layer.
There are also many different gateways (appliances, software or VSAs) that in addition to providing NAS access, can do iSCSI among other means of attaching to an object storage repository (product or service). Some of the products such as DDN WOS, EMC ATMOS among others have also started adding IOS among other interfaces in addition to REST/API/S3 or NFS, CIFS, FUSE or programatic bindings for flexibility, then again, others have rather limited access. Likewise some do a better job of encrypting on the source side as well as single instancing, compressing and deduping before sending vs. others.
So is NAS or file along with block dead at the hands of object, maybe for some or perhaps wishful thinking for others, however they are still very much alive, granted their role and location are changing. You can have object storage (product or service) accessed via NAS (gateway, software product, etc) that in turn has its object software stack (ATMOS, Basho Riak CS, Ceph or Cleversafe, Dell DX/Caringo, OpenStack Swift, Scalaity among others) that in turn sits ontop a lower level filesystem (btrfs, etc.) that is not seen by users, that in turn sits over block storage devices (not seen by the objects).
Hope all is well, cheers gs @storageio
Greg,
thank you for the comment,
I’ve (over)simplified the concept, obviosuly NAS isn’t dead. 😉
My point here is that a modern storage system needs more access protocols than in the past, especially now that mobile devices (iPads) are playing an important role in enterprises. IMHO, some vendors are working on object storage as a backend because I think it’s easier to maintain consistency of the whole system if the same data are accessed through multiple protocols/devices
best,
Enrico
No worries Enrico and simplifying object storage will help to remove one of the many obsticles to its adoption. Adding more types of access can be done independent of object storage per say, for example it is possible to add other interfaces, protocols and personalities to NAS as some have done where you can have NFS, CIFS, AFP, propritary, block, HDFS, HTTP among others. Granted some interfaces and methods may be easier with some object based systems/services vs. others, and some object systems or services will also lag in supporting different access methods.
To your point, some of the vendors and services are now adding IOS as a means of access storage among othes, in addition to NFS, CIFS, S3, etc. If you have not done so lately, take a look at the long list of interfaces (native and via gateways or 3rd parties) for DDN WOS, EMC ATMOS and some of the others. Some are standard or familar methods, some not so familar depending on your area of experience.
Again to your point, the major obsticle that is facing the current generation of object storage systems (as opposed to object archtectures which is a different discussion) is the same as those from the past generation (e.g. Centera, Caringo, HCP/Arkivas, RISS (ask HP where is it now), Bycast(NetApp Storegrid) among others) is ease of access, interoperabity with existing applications and environments.
As with the last generation of object storage based systems/solutions, some did well with adding API and application support, native and external gateways or bridges or what ever was needed to make the solution plug into an existing environment, application ecosystem and do its job. Otoh, some took a different approach focused more on selling the virtues of the technology, its architecture, paradigm shifts, buzzword bingo, etc. and missed the business message and oppourtunity.
Thus, who ever can make thier solution or service or stack work with existing applications, devices, access methods along with new and emerging ones will have success while others complain about those having success. Sound familar?
Cheers gs @storageio
Enrico and Greg
Although the article as to be taken with a grain of salt (death of any technology is always a long drawn out process)…your discussion in the comments is excellent.
My 2 cents:
The future of NAS is the “Buy a NAS get a CLOUD” principle”. By aggregating “Loosely connected & distributed” building bricks, each supporting a multi-protocol front end on top of a collaborative object store, we can build highly scalable, application friendly and collaborative store infrastructure. Being protocol agnostic will ensure interoperability and ease of access while a well designed object back end ensure resiliency & performance. I also provide a bridge for thousands of existing legacy applications that cannot be refactored overnight.
Hi Enrico,
The headline about NAS may be overstating the trend but I agree with the sentiment, especially in the context of private file sync & share solutions. We at CTERA have several such deployments using EMC Atmos as the backend, including Swisscom’s Storebox service, another service that will be announced by a major European telco in a few weeks, and a large European bank in a private deployment that already deployed to thousands of end-users.
Some of these also combine (or plan to combine) cloud storage gateways as part of the overall architecture to optimize user experience in branch offices and remote locations.
What we found to be the key criteria (which you correctly pointed out in a previous post) are deep, seamless Active Directory integration, support for private object storage backend, and completely private encryption key management. Plus of course it has to scale, provide HA and redundancy capabilities, and support the popular platforms (iOS, Android, Windows, Mac, Linux).
Rani@ctera.com