EMC has serious rationalization problems now; Centera vs Atmos vs Mozy/Decho. Centera has the disk archival market and OEM channels wrapped up, and Mozy/Decho is a cloud service already in production with a jillion TB under management (nice design too!).

…all of which is largely irrelevant, of course, because Joe Kraska is right: Atmos is not the same as scale-out NAS. It’s more similar to CAS, in some ways, but the real point is using simpler APIs and looser consistency (compared to a “true” filesystem) to support large scale and global distribution. As anyone who has worked with parallel or distributed filesystems is painfully aware, even NFS’s weakened-POSIX semantics can get in the way of providing maximum performance and resilience per dollar, and unloading some of the baggage can be very good for applications that didn’t need it anyway. For examples, look at what the PVFS2 or GoogleFS guys have to say about POSIX compliance, or the rationale given in Amazon’s SOSP paper on Dynamo. That’s what internet scale looks like, and it doesn’t look like any kind of NAS including clustered NAS.

There are two issues:
1. this is a niche product. Not every company is going to be able to use Atmos. Those that do will have a need for huge amount of the storage.

2. EMC’s sales force is not tied into the right people in those accounts yet. EMC is used to talking to people with block IO and relatively high performance need. It’s going to take time to foster those relationships.

I agree that there are challenges ahead, but sales force hostility is not one of them.

I don’t view Atmos as competing with Isilon or GX at all. Atmos is a commodity hardware scale-out architecture for the web crowd. It has Akamai-like functionality built into it with a global namespace built into it, thus it is able to handle distribution of- and transparent access to- files around the globe. Amongst other things, this includes the ability to transparently fail over access to files in your enterprise in the event an entire data center is lost. Neither GX nor Isilon can do this.

If you engage the Atmos team, they will tell you clearly that this is what Atmos is for. They really designed from scratch for handling a geographically-dispersed storage enterprise. You have to give them kudos for this; it’s big thinking.

The demand for this sort of product is a very narrow niche, but the volumes in this niche are very high. I expect that what they will do is sell small numbers of accounts that by very large volumes of storage. If you’d seen the pricing on Atmos, you’d be simply floored.

BTW, if for some reason you don’t like EMC, you can get much of Atmos’ functionality by way of ByCast StorageGRID, which is 10 years old with much proven functionality. Caveat, it appears to be more expensive.

Joe.

The first case I saw this happen was when the company I worked for introduced a lower cost workstation for the sub $10,000 range. Sales did not jump on it because they were quite happy selling the $35K to $50K versions. Net result was sales which were supposed to be in the hundreds of thousands was only in the thousands. The product died a slow painful death.

Second company I worked at where a new product would potentially compete with the old more expensive versions management realized up front that the sales and marketing teams would not be on board. So as soon as the prototypes were finished and working the development group was spun out into a separate company. Net result flourishing and profitable young company outselling the old company and its competitors. Shareholders are happy because they got more value out of the spin-off than they would have from the parent company’s regular growth pattern.

Terry

The problem is that EMC is like HP w/ the 9000- probably 3 years late to the party- this market is already dominated by Isilon and Spinnaker/NetApp. NAS and Scale-out NAS is not their wheelhouse- NetApp invented NAS.

While the analyists/bloggers loved it, Atmos/Hulk’s announcement was greeted by a collective ‘who cares’ from the people who actually matter- the companies who buy the technology. The response I saw was one of disappointment- companies were expecting scale-out block storage (XIV/Left Hand) and got scale-out NAS.