Some days, I just feel like I’m out of the CTO loop. I see announcements for new “Web 2.0 Cloud Optimized Storage” and my brain gets all foggy trying to see through the new goofy buzzwords and trying to figure out what it’s even for in the first place. Now, if you’ve met me then you know I’m a pretty cynical guy… although I personally prefer the term “pragmatic”. Perhaps I need to go back for some remedial classes in being a public commentator?
I’m talking about EMC’s Atmos, of course, née Maui. Last week EMC broke with a tradition of not announcing new products in economic downturns, and launched a peculiar new storage product that’s neither fish nor fowl. Oh, it’s got the requisite NFS and CIFS interfaces, but those are surely the unloved stepchildren like the Centera Universal Access Gateway was; a way to make you feel more comfortable with the new device but not a full-fledged, first class interface. Instead it’s half content distribution network, half object storage, and all rather weird.
From what I can find, Atmos is a new object storage device, with a new proprietary interface. (Don’t let those SOAP and REST acronyms fool you — it’s still a proprietary interface. SOAP provides a session layer, but the requests are all Atmos-specific. Claiming any given protocol as standard just because it uses SOAP is about the same as claiming it as standard because it runs over TCP/IP. Or uses ones and zeros.) It comes in three sizes, rather big (120 TB), rather bigger (240 TB) and rather biggest (360 TB). It has no in-built data protection, so you need to have multiple racks in multiple sites from the get-go, and they better have really fast connectivity between them. And it has rudimentary policy tags for object metadata.
So, it sounds like everyone who rewrote their applications to work with the Centera gets to rewrite their application again to work with a new EMC-specific interface. They didn’t even choose to use XAM, an object interface standard we all worked on to deliver rich metadata objects with a global namespace, and a standard capable of expressing all that Atmos claims to do and far, far more. Strange choice?
Steve Todd is very excited, but I can’t blame him, it’s his job. He hasn’t convinced me to be excited about it yet. Web 2.0 is just a vague term that Tim O’Reilly made popular, and it’s about interactive, network-centric applications in a general sense. EBay is Web 2.0. Google Maps is Web 2.0. Heck, you can comment here and link through a syndicated identity structure, so this very blog is Web 2.0! None of it has to do with storage beyond the fact that all these web applications need storage, and Steve hasn’t convinced me that a brand new simplistic object interface with three racks of equipment is any better for storing my blog posts than a ZFS file system.
Content distribution and content replication is potentially neat, although we’ve had WAFS technologies out there for file systems for some time. I’m curious to find out what sort of locking, transactional consistency, and disconnected operation Atmos allows for, so the jury is still out here.
However, Robin is pretty excited about Atmos too, and it’s not his job. He likes that EMC is striking out with a new way of doing things instead of waiting for someone else to come along and eat up their market share. Best to jump first and introduce customers to something new yourself — this worked well for EMC with Centera. (On the other hand, you might want to ask the Centera customers how well it worked for them.) That’s a fair point.
Last week I did say that now is the time to consider investing in new, cheaper, more effective storage architectures so that you can save money in tough times. I definitely stand by that, but I’m not saying you should change your whole application chain to fit into totally new interfaces at the same time….
Looking around more, Tony Pearson is also really excited. He’s enumerated his reasons in a Thanksgiving post!
Most of all, he’s thrilled that there’s no longer a stalking horse Maui preventing customers from considering IBM’s products. Now that Atmos is out there, customers can make a decision on if it’s right for them. If not, they don’t have to worry about missing out and can buy something else. That’s an extremely good thing for IBM, who also has products (XIV) that only start at really, really big sizes. He also likes the other bad things about Atmos, because they also match XIV — big capacity increments, lack of powerful, low-overhead data protection (replication only?), and Web 2.0 buzz.
If I was at IBM, I’d be convinced too. This is great news for them…. but I’m just not seeing how it’s great news for the storage consumer!
I’m going to be a grouch here and declare a moratorium on new storage terminology without a majority vote. Cloud storage used to mean Storage as a Service (StaaS, another horrid acronym); things like Amazon S3 or the internal, special purpose clusters that people like Google have developed. Now it’s expanded to object-based content distribution networks. Each product is in a class by itself, and that’s not a good thing.
All the storage bloggers have catchy names; we’ve got a storage anarchist, storage architects, and storage insights. Maybe I should be the storage sourpuss?