With today's announcement, the uniqueness of the category becomes more apparent. We're now in a better position to make the case that -- for some use cases -- cloud-optimized storage solutions can be better than their traditional counterparts.
What Was Announced
If you're not familiar with Atmos, you're not alone. It doesn't fit neatly into the SAN/NAS/DAS categories that we're all familiar with. Competitors, in particular, don't really know what to make of it.
Mutliple Atmos units work together to create a storage cloud. All information is stored in objects with integrated rich metadata. Storage policy (how many copies, protection method, physical location of data, spin down, compression, etc.) is expressed in terms of service level desired -- gold, sliver, bronze, etc. Atmos interprets the policy intentions and seamless orchestrates all the data logistics in the background. A single management console monitors service delivery.
It takes a while to fully comprehend what Atmos really does, but -- regardless of your vendor affiliations -- there's no denying that Atmos is a unique solution to a unique problem -- overcoming, and then ultimately exploiting -- distance.The Atmos Loyalists
Not every enterprise needs to run their own storage cloud. That being said, you'd be surprised how many enterprises are interested in cloud storage services powered by Atmos.
The organizations that have stood up significant Atmos storage clouds are extremely pleased with the results, which is gratifying. They've told us that there is nothing in the market that's remotely close for this use case. That's also nice to hear.
Some of these organizations are service providers, and that's where it gets interesting.
Since Atmos is designed to be inherently multi-tenant (built in, not bolted-on), it can be the basis of a rich set of cloud storage services that are hard to match (either economically or feature-wise) using traditional approaches.Examples abound: high-performance content delivery, archival records storage, even the tail-end of a backup routine where the final step in the hierarchy is a cost-effective -- and extremely resilient -- cloud storage service.
Where Cloud Storage Can Be Better
Understandably, most people think "cheap" or "elastic" when they think "cloud" -- but there's more to it than that, especially when it comes to information storage.
As I outlined in my post "Exploiting Distance", information in the cloud can take on some unique properties. In addition to exploiting economics, cloud-based approaches can also deliver better user experiences (by locating the information closer to the user) as well as substantially better forms of information protection (by dispersing information in multiple locations.
Which brings us to the new Atmos feature: GeoProtect
Parity Protection In The Cloud
On the surface of things, most storage people will quickly recognize what GeoProtect does -- it's parity protection (or, more precisely, erasure encoding) across multiple geographically dispersed nodes.
Choose the number of data segments you'd like, and the number of parity segments you'd like. Information is transparently distributed using the existing Atmos policy mechanisms.
Previous versions of Atmos storage software typically had to keep a separate physical copy on a remote node in case the primary failed (RAID 1?); now there's a rich variety of parity schemes available, 5+2, 7+3, and so on.
Sure, this is more cost-effective for storage media than before, but if that's what you're focusing on, I think you're missing the big benefit -- a new level of protection that we really haven't been able to consider before *and* at a lower cost than before.
Consider This, Please ...
Imagine a average-sized Atmos network of 9 nodes, all in different time zones around the world. And imagine that we were using, say, a 6+3 protection scheme.
The implication is clear: any 3 nodes could be completely lost: failed, destroyed, seized by the government, etc. -- and the information could be completely recovered from the surviving nodes.
Now that's redundancy!
For organizations worried about their information falling into the wrong hands (whether criminal or government sponsored!), any subset of the nodes would yield nothing of value -- not only would the information be presumably encrypted, but only a few slices of a far bigger picture would be lost.
Here Comes The Bold Claim
I believe that Atmos using the GeoProtect feature offers a level of information protection unmatched by any commercial technology today, and does so at a cost far below many competitive alternatives.
Faster, better, cheaper. That's a combination that's hard to beat ...
A Platform, Not A Feature
What makes this even more attractive in my mind is that this new feature is a simple extension of the existing data strategies that Atmos storage can support, all orchestrated using the same generic (and powerful) policy and object abstractions.
A given piece of content might have multiple physical copies made automatically if a great user experience is desired, or be automatically moved to low-cost storage, compressed and spun-down to save money, or -- now -- automatically geographically dispersed to dramatically increase protection.
All done with a minimum of fuss. Build your own Atmos cloud storage environment, or rent one from a growing number of service providers. EMC has also put up a limited implementation of the service for others to evaluate, which you can find here.
What Comes Next?
Despite my obvious vendor biases, I think it's fair to say there's an entirely new set of options now available to understand and consider. Once we get past the inevitable market lag, I think many organization will realize that a significant portion of their information portfolio will be a fit for something like cloud-optimized storage -- ideally powered by Atmos storage.
I also thinks this puts an entirely new wrinkle on the overall industry "cloud" debate. It's one thing to debate internal vs. external economics. It's another thing entirely when the external service provides attractive capabilities that just aren't practical with internal approaches.It's funny -- for the last 20 years, the storage industry has been powered by the idea of RAID in the array.
Will the next 20 years be powered by the idea of RAID in the cloud?