EMC Atmos has always been rather difficult for traditional storage people to fully appreciate, but fortunately we've had no such problem with the newer generation of web developers and service providers :)
In a nutshell, Atmos software is used to create cloud storage services that are a sharp departure from what's normally available from, say, Amazon's S3.
And if you're in the business of delivering storage-based services -- or work for a company that needs to do this for its own purposes -- it's not hard to have an appreciation for what Atmos does -- and how it's quite different than just about anything out there.
Atmos was first introduced in November, 2008. Since then, it's enjoyed steady adoption from a growing number of customers and partners who now have an deep appreciation of what does better than anything out there.
Atmos By The Numbers
The wide geographic ability of Atmos storage services can be an advantage to more centralized S3-like services both in terms of customer experience (data closer to consumer) as well as being more compliant with data location regulations.
There are some seriously big Atmos clouds in production, including one whopper with over 100 petabytes scattered over 155 nodes, supporting many billions of objects. At some point, I'd love to total up all the aggregate Atmos capacity and objects, and compare them to Amazon's.
I wouldn't be surprised if we end up with a bigger number.
One of the topics that takes a while for people to get their heads wrapped around are all the different consumption models available, or the fact that they can easily be combined.
If you're interested in Atmos, you can:
- buy your own physical nodes and run them as your own service
- buy Atmos software in VMs and run them on your own hardware
- contract with a service provider for a hosted Atmos service that you manage and control
- simply consume the service via APIs for both data access and certain management functions
The functionality characteristics of Atmos are identical regardless of your choice in consumption models. One of the many advantages of this approach is that it allows global cloud storage services to be more easily created without a corresponding global investment :)
Another difficult concept for some to grasp is that Atmos is a metadata-driven storage model.
Metadata -- bound to the object -- drives every aspect of Atmos storage cloud behavior: cost, performance, protection, location, security, etc. This metadata model is also one of the foundations around a very rich and sophisticated multi-tenancy model: tenants, tenants of tenants, tenants of tenants of tenants, and so on.
Finally, if you tend to think of all your storage living in one data center, Atmos will be hard to appreciate. In a dispersed Atmos environment, information is always on the move from where it lives to where it needs to be: for performance reasons, cost, protection, etc. Yes, it manages as a single image, but one with global scale.
Simply put, interested parties can use Atmos to quickly stand up global cloud storage services that aren't available elsewhere in the market. And it's turning out to be a really big deal for some.
What's New In Atmos
The big news is the new "Sumatra" software release, available later this year.
Atmos performance has historically been network-bound: there are a few new interesting network configuration options that allow more bandwidth into (and out of) dispersed Atmos nodes.
There's an improved event manager workflow, the ability to aggregate logs to feed into higher-level analytical tools, and related enhancements. Atmos upgrades have always been non-disruptive in that they transparently fail over to a redundant node during the upgrade: this window has been dramatically reduced using a "fast reboot" process to minimize impact even further.
Making Atmos Applications Even Easier To Build
For example, Atmos has supported a Firefox browser plug-in for a while, now there's a Chrome addition. There's been a great wrapper for iOS applications, now also available for Android.
There's now an interesting and surprisingly useful token management system that provides very fine-grained control over who can upload and download objects.
Many applications are coming over from the Centera world which had its own rich metadata: Atmos has always preserved that metadata, but now has more ability to use and act on it to drive more policy behavior.
And there's a new open-source application (AtmosSync) for people who are looking to move piles of data into (or out of) an Atmos environment, usually a NAS mount point or -- more frequently -- an Amazon environment. Not announced here (but definitely being worked on) is support for the Amazon storage APIs against an Atmos storage cloud -- another choice for app developers.
Great Stories Abound
More interesting -- and more generally applicable -- is what can be done in the hands of smart IT service providers: NIS will be talking about how they use Atmos and its development tools to stand up all sorts of storage-based services quickly, easily and cost-effectively.
They were a bit skeptical at first (as most people are), but now they're believers :)
Plenty Of Great Links If You're Interested
The team has done a great job making all sorts of interesting resources available for developers over the web.
For example, if you're curious what it takes to write a simple Atmos application, you can try it for yourself here. If you want a quick, 2-minute overview of what it does, try this link here. A great developer community here. A nice blog here. And, of course, tons of other resources on their home page here.
Why rent someone else's limited storage cloud -- when you can (virtually) build your own?