I spent most of last week at Cloud Computing Expo, attending, presenting but mostly exhibiting with our partner Panzura.
I spend a lot of time on shows like this (From InterOp to HostingCon, to Storage Network World) and CCE always struck me as being the one where we needed to educate the attendees the most. I don’t know why, but that’s the experience I’ve had in the many CCE I’ve been to, either as attendee or exhibitor. However, last week was a different, it was the first time that I felt I did less educating and more pitching. And in a way, I’m not sure I like it better;)
Object Storage is not a hipster thing anymore
In the past 2-3 years since we started Scality, we have had to educate and evangelize about Object Storage a lot. Whereas Cloud Computing is now experiencing a pretty comfortable understanding and adoption (with still some caveats and work needed for massive, broad adoption, of course), Object Storage still had a bad reputation of unreliable, low-performance storage technology not suitable for large-scale and/or enterprise-class storage, up until late last year.
What I’ve been seeing in the past 8 months is that first, Analysts have finally recognized the role Object Storage will have in the Enterprise, and second, more recently and even more specifically at this show, Enterprises themselves are finally shifting towards a better understanding of this new tech. I’ve seen this particularly at this show because since it was in NY, a lot of the attendance were financial institutions, banks, and government contractors who came up from Maryland, Virginia or DC. They usually always came up to us with questions on how Object Stores could do some of the specific features of SANs or NAS (how do I manage backups, for example). Last week, I experienced some very different questions about helping them solve issues that SANs and NAS couldn’t anymore.
Of course, Service Providers and Telcos also had the same questions earlier than last year, but they came aboard the Object Storage ship earlier because they have come across the scale problem before most enterprises. They also were not very present on the show floor last week, so I’m going to focus on what I’ve experienced last week.
There is still something that, incredibly, still needs education
As incredible as it may seem, i realized early something during the show and some conversations I had the week before. Even though Object Storage might now come through as a legitimate Storage technology, what to use it for is still confusing for a lot of people.
I had a presentation planned on Wednesday morning at 815am during the show, again with our Partners Panzura. I decided to focus my part of the presentation on what and why Object Storage is for.
And it all comes down to Structure
Object Storage is made for Unstructured Data, and that’s what baffles a lot of people it seems. During my presentation, i made a point of explaining as easily as possible what Structured and Unstructured data are. You can find my slides here.
It really boils down to this:
- Structured data is data that needs an external structure to be made relevant. Look at a database row. Taken alone, a cell makes no sense. It has to be linked to another piece of data: how many people are born on the same day? without a name or an SSN attached to it, 12/8/1976 doesn’t mean anything.
- Unstructured data is a self-contained file, relevant in and of itself. Think about a Powerpoint file, it has content, metadata (author, revision, date of creation) accessible within one file alone.
I’ve used it to explain the difference on the floor, and it was pretty successful. Once people understand this, they are less likely to ask about whether or not we can store hot VM images or Relational Databases, and we can have a more fruitful discussion.
The other thing that became even clearer last week, and which is one of the reason Enterprises are starting to look for other solutions than SANs and NASs is that they’re reaching a point in data storage and growth that is just not sustainable anymore with those solutions. I’ve talked to some people already storing multiple PBs and some rarer ones who are generating an insane amount of data (hundreds of TB a day, as per their claims). At this scale, one needs to think further than the SANs and NAS that are sprawling in their Datacenters.
I’ve heard today that NetApp announced at their Summit that they have a customer who deployed 750 PB of storage, with hundreds of other 10PB deployments. Those are impressive numbers, but I would love to hear details about those deployments in terms of cost, size of the ops teams and migration plans for the next technology refresh. I am convinced that those 750PBs are, in fact, hundreds of “small”(er) deployments and not one flat namespace with one single storage infrastructure for multiple services, as an Object Storage like Scality can do.
Which makes the number less impressive. But still impressive nonetheless in terms of revenues, I’m sure.
Amazon has Exabytes of data using (their own) Object Storage technology and commodity hardware and that is where we’re headed for the Unstructured data of this world. Remember that IDC considers that Humanity will generate 2.7 ZB of data in 2012. That’s 2.7 Billion latest generation high-end MacBook Pros. At that scale, there are no other options than Object Storage.