Since 2006, when Amazon introduced its Simple Storage Service (S3), organizations concerned about security and service levels have been eagerly waiting for a viable marketplace of storage services to emerge. The industry lately has been doing sort of a replay of 1998-2001 when storage service providers were all the rage. Those who are old enough to remember Storage Networks, Inc. will recall that the company was the hot storage startup of the day, threatening the business models of the established enterprise players.
Then SNI imploded. Realizing it had no viable business model to pursue, the company decided to shut down and return cash on the balance sheet to investors before the market wised up to the company’s flawed model.
On paper SNI was brilliant – outsource storage to a service provider with points of presence all over the world. Pay on a usage basis and let the service provider bear the CAPEX cost of infrastructure, allowing customers to pay only for what they consume. Of course the problem was that much of the world’s data at the time was structured content (e.g. databases) and needed to be proximate to the server (nagging speed of light issue) making latencies onerous in the SNI model. Combine this with the fact that there was really no multi-tenancy capability in place (squashing leverage and buying power) and that SNI ultimately had a more expensive cost model than on-premise storage (even for use cases like backup and archiving) and voila – the bubble burst.
What’s Different Now?
Fast forward to today where the world’s content is predominantly unstructured and there are far more applications where performance is less of an issue. More bandwidth means lower overall communications costs and the concept of multi-tenancy is actually more than a concept for some service providers. It means that small and mid-sized organizations can actually access the same types of infrastructure as the largest financial companies without the need for expensive IT personnel and up front infrastructure costs.
In the past few years we’ve seen VC money pour back into the cloud storage space, which has been encouraging to buyers looking for alternatives to S3. There’s nothing wrong with S3 mind you – we use it here at Wikibon, but many organizations either don’t have the developer skills to leverage it or don’t trust Amazon’s lightweight SLAs. [Note: recently Amazon has introduced platinum level service but it’s expensive at $180K annually].
Despite all the promise, we’ve seen a mini implosion recently in this space. Last week, a well-funded company, Cirtas blew up. In a news story that was broken by SiliconAngle’s John Furrier last Friday, we found out the company’s product essentially doesn’t work and is laying off a bunch of people. Cirtas is buildng a cloud onramp, basically a piece of hardware that sits on your data center floor and is supposed to move data into the cloud storage network of your choice. VCs, a bit too eager to invest in anything associated with the word cloud these days, pumped over $20M into the startup in the last three months—and over $30M to date. But the value from an infrastructure standpoint is the actual cloud storage repository. As we’ve pointed out before:
All the major storage vendors need a platform that offers usage-based billing, secure multi-tenancy, a large scale (i.e. billion plus) object store, support for cloud protocols (e.g. REST and SOAP), a global namespace, proven data integrity and an open set of APIs that allows easy entries and exits into and out of the platform to facilitate integration. These are the core capabilities that will allow the major storage vendors and their ITOs to deliver Amazon-like capabilities and at the same time provide enterprise-class services that can drive premium value for their organizations.
Cirtas was a gateway to get to a cloud repository that performed the above functions—it didn’t store customers’ data itself. And in today’s market, customers are looking for pay-by-the-drink storage, not more up-front hardware costs. The fact is that no one was buying Cirtas’ solution – it just wasn’t selling. Moreover, based on our checks, it had fundamental hardware performance and code issues as well.
So the bottom line is VCs looked at a product that wasn’t working, a direct sales model that didn’t take off, and little to no customer interest in Cirtas as an enterprise value-add feature, and effectively made a decision to stop chasing windmills. It determined the best course of action was to pull back the spending, fix the product code as best as possible and perhaps sell to any (potentially) interested parties.
The problem here is a cloud on-ramp is basically a feature. And while sometimes so-called feature companies can excel (3PAR, Data Domain) it’s more often the case that the successful ones actually have an architecture that works and is disruptive.
Last week’s Cirtas “do over” should give standalone cloud on-ramp companies a wake up call. The company most at risk, and with a similar model to Cirtas, is Nasuni, which is a lower-end, SMB-grade gateway which like many new companies is experiencing growing pains of their own in some customer sites. Cirtas tried to focus on the enterprise and Nasuni tries to focus on the SMB space, but both are really under-developed products searching for a business model.
The fact is, to succeed as a cloud gateway, you have to be part of a bigger, growing business or you have to offer extreme levels of competitive differentiation. Case in point, look at Riverbed. Riverbed has diversity. It has a $5B market cap and meaningful revenue ($500M+) and is successfully leveraging WAN acceleration expertise and IP to develop highly differentiated, fast cloud gateways—its whitewater appliances—which are targeted directly at an enterprise customer base. Riverbed has thousands of customers it already sells WAN accelerators to a captive base that it can now go sell Whitewater appliances to as well—creating a full blown product upsell cycle that leaves little room for startups with no differentiation. In my view, Riverbed realizes that cloud gateways are a feature, not a standalone business.
That’s not to say firms can focus on the so-called feature strategy. As far as differentiation goes, take a look at Panzura. The Panzura Alto Cloud Controllers are very fast, offering line speeds that match those of public cloud storage networks, and are feature heavy—with built-in support for CIFS, NFS, File Locking, Versioning, Snapshots, Compression, Dedupe, Cache—and the product supposedly works. It’s not a lab experiment that’s not ready for prime time and it’s not a low-end, scalability-challenged box with little more than a cool name going for it.
Iron Mountain Packs up its Cloud
So that’s my rant on cloud on ramps. Also last week we saw the collapse of Iron Mountain Digital’s cloud storage business. Steve Duplessie put it best when he compared Iron Mountain with Amazon:
Amazon has a great model. You pay what they want you to pay, which is so cheap that you couldn’t possibly do it yourself–unless you absolutely, positively care about accessing your data (securely)–in YOUR timeframe. As long as you don’t care, Amazon’s scale makes it almost impossible for it to lose money–it’s really just a matter of how much it can make. It’s a great business.
He also commented on Nirvanix:
Nirvanix believes it has some operating advantage–i.e., it can store, replicate, manage, and deliver your data in its cloud cheaper, faster, and better than anyone else can (or than you can on your own). If it does, it can make money as soon as it crosses whatever elasticity point exists in the business (i.e., once we get 10 PB, our cost per TB stored is decreased 90% … ). Then it gets economies of scale and can make money as long as it keeps customers happy. It’s a great model.
Both Amazon and Nirvanix either have operating leverage that Iron Mountain doesn’t have or the return on that leverage isn’t great enough compared to other areas of investment for the company.
Unlike Amazon and Nirvanix, Iron Mountain is still valued on its ability to return on its assets–perhaps the lowest tech means of valuing a company, but it is both its blessing and its curse. Iron Mountain gets a higher return on its debt/assets in its traditional businesses and its high-value add digital businesses than they can ever hope to get on selling capacity. It’s as simple as that.
The point is Iron Mountain didn’t have the stomach or the wherewithal to compete with Amazon.
So at the end of the day, one vendor couldn’t succeed with a standalone feature (Cirtas). The other couldn’t succeed with a me-too play in a world completely foreign to them (Iron Mountain).
What’s the Future Hold?
And what will happen to the cloud storage providers like Nirvanix and Zetta? Nirvanix had some bumps in the road and had to re-tool with new management – but is now getting traction with real customers and partnerships. There’s clear momentum there and the key is getting sales to the volume point of profitability.
How about Zetta, which primarily does server-to-server replication– like a Double-Take or a CA XOsoft, but at the volume end of the market. Initially Zetta went after the market for primary cloud storage, but found out that nobody wanted to store primary, structured data in the cloud (hello SNI). Can the company gain market traction? It’s unclear at this point if Zetta’s strategy is working but word is installs peak out at around the 30TB range. Is that enough to hit profitability? If the company can sell to a large number of customers in volume yes, if not, no.
At the end of the day, the final battle for a major piece of the action in the cloud will be fought amongst the big whales—Dell, EMC, HP, IBM, Hitachi and Oracle—vs. the Kings of the Web—Amazon, Google and Microsoft. The key thing to watch is who gets all the right pieces of artillery—the features, the software, the IP, to add to their cloud portfolio. Moves will be made, titans will clash, and the market for cloud will continue to grow at an accelerated pace
HP and Dell are ones to watch closely on the acquisition front as both companies are re-inventing themselves. HP’s Leo Apotheker has announced the company will enter the public cloud market this year. Will it do so through an acquisition or build its own public cloud? HP or Dell might for example take out a Nirvanix or a Zetta to get the IP necessary to compete with Amazon on the public cloud front quickly.
Despite the recent fallouts, the enterprise cloud market is maturing and just starting to hit its stride. This is good news for buyers as it is highly disruptive to traditional on-premise storage models, which are way too expensive and can dramatically improve applications such as backup, archiving and disaster recovery that are screaming to be outsourced.