I've been writing about hybrid cloud models recently, making the case for extending cloud infrastructure on-premise in certain circumstances. I did so with foreknowledge of an announcement that BPM vendor Intalio has had in the works for several months. Last Tuesday, I was in Palo Alto for the public unveiling of Intalio's new application platform, which is available both on-demand and on-premise — but in a model that reverses the normal polarity of such offerings and challenges the received wisdoms of cloud purists and on-premise diehards alike. It does this by taking some of the core principles of multi-tenancy and moving them on-premise. Read James Taylor for a detailed and thorough write-up of the product announcement. Perhaps most surprising, this challenge to on-demand orthodoxy is led by Intalio's founder and CEO Ismael Ghalimi, a long-term evangelist for the on-demand model and organiser of the annual Office 2.0 conference, a mecca for advocates of on-demand computing.
As I've discussed in previous posts, there are two equally valid ways of achieving multi-tenancy. The first, espoused and evangelized by Salesforce.com, is to share as many tenants as physically possible on a single instance, right down to and including the database layer. The second model, adopted by NetSuite, SuccessFactors and many other leading SaaS players, is to replicate instances across large numbers of low-cost commodity hardware machines, sharing databases for smaller customers but having larger accounts run on their own virtual database instance. This is still multi-tenancy because the replicated database schemas are identical. There is nothing that ties any customer to a specific instance, leaving the provider free to distribute and manage instances in whatever way it sees fit.
What Intalio has done is to borrow this principle of replicated instances and apply it to its on-premise implementations, as Ghalimi explained last Tuesday:
"One of the lessons that we've learned in the past 20 years is that letting customers change application schemas is a very bad idea. We've learned that lesson and the way we implemented our CRM application is, you cannot change the standard objects. All you can do is expand them. You can add fields, but you can't remove them. You can rename a field but you can't change its logical name. So any modifications customers do will not break the schema. We've learned a lot from Salesforce.com. They did a really great job there."
This is a perfect illustration of a principle that — in opposition to Microsoft's 'Software-plus-Services' rhetoric — I prefer to call services-plus-software. In other words, architect for the cloud first, and then (if you must) put the software on-premise as an extension of the cloud rather than as a discrete instance. In adopting this model, Intalio is taking to its logical conclusion a shift that mainstream vendors including SAP and Oracle have already begun. They've recognized how much endless customizations have hampered rather than helped their customers and are moving towards more 'cookie-cutter' implementations of the core application stack and what SAP's CEO Leo Apotheker has called "innovation without scary upgrades and sleepless nights." But the ingrained habits of customers and the business models of integration partners make it impossible for a company like SAP to adopt the purist, schema-intacta philosophy that Intalio has espoused. That's the challenge Intalio poses to on-premise vendors.
Cloud purists face a starker challenge in Intalio's insistence on enabling on-premise implementations of its replicated-instance cloud platform:
"We think on-demand is great, but we think choice is even better," said Ghalimi last week. "We think putting this arbitrary limit on how the product can be deployed is wrong."
Intalio's platform — which includes a CRM application that it claims is functionally equivalent to Salesforce.com's unlimited edition, as well as a business process management system with a user-friendly visual design interface — is offered in three deployment modes, all licensed on a subscription basis. As well as on-demand and on-premise versions, there's a managed on-premise option, in which Intalio provides the hardware, software and operational support, while the customer provides the power, cooling and bandwidth. This version is designed to appeal to larger enterprises, with a 2,000-user minimum and a three-year contract, but delivered with all the deferred-cost and time-to-market benefits of the on-demand model:
"It only comes from an op-ex budget, there is no cap-ex [capital expenditure]. We bring everything, we manage it for you," said Ghalimi. "You can call us today and in less than two weeks you'll get an appliance that's fully managed by our people and has BPM and CRM running."
The model removes the biggest obstacle for customers, he explained. "The issue is procuring the hardware — getting a capex budget is very hard today — installing and managing this hardware, especially if it's a little bit fancy, and then managing the software stack on top of that. That's really what we take care of with the managed on-premise option. It's very much an on-demand system that just so happens to run in your on-premise data center ... So far, every single customer who said we want the cloud, went for the managed on-premise option."
There's still room to debate the relative merits of a provider having all its customer's instances in the same data center infrastructure rather than spread around a large number of different sites. But by narrowly defining its stack — not just the core application objects but also the specific application server, database and Linux version, even the hardware platform in the managed on-premise model — Intalio has brought its support and management overhead so close to a provider-hosted model as to be almost indistinguishable. Indeed, Ghalimi even claims that Intalio can operate a "Salesforce.com-class cloud" — 55,000 customer accounts and 1.5 million user seats with multiple GB of storage per customer — at an operating cost of less than half a million dollars per month, or less than half of one percent of Salesforce.com's revenue.
Whatever marginal operating cost advantage Intalio may lose by moving on-premise is more than outweighed by savings the customer makes, for example by using its own in-house bandwidth, or easing integration to existing on-premise assets. Indeed, Intalio's very first managed on-premise customer plans to consolidate its SAP system onto the same Intalio-owned blade servers, reducing its operating costs while facilitating integration and data exchange between the two systems.
From the vendor's point of view, of course, implementing on-premise also removes many of the objections a venture-funded start-up will face when selling into the large enterprise market that Intalio is targeting (although its on-demand model starts at three users paying $9 each per month, its sales efforts and existing customer base are concentrated in the 2000+ seat range). It also helps that Intalio has adopted an open-source model, giving customers full access to the source code of its platform. Another comfort factor comes from the ability to run some or all the platform on public clouds such as Amazon EC2 or Google AppEngine (at the time of writing, Intalio is running its own on-demand implementation on EC2, pending final readiness of the hardware for its own data center).
Abandoning an on-demand only stance will appeal to systems integrators and ISVs, too, Ghalimi believes (a similar bet to that recently made by LongJump, another PaaS provider):
"One of the main problems for the on-demand model of [Salesforce.com's platform-as-a-service] Force.com is that it's not very appealing for software vendors," said Ghalimi. "The lack of an on-premise option is something that's limiting the community of ISVs on Force.com."
But despite making so many comparisons to Salesforce.com, Ghalimi told me Intalio isn't aiming to take market share from the SaaS powerhouse.
"We have tremendous respect for Salesforce.com — they really created cloud computing," he said. "I don't think we compete against Salesforce.com. We sell very similar features. The main difference is, even though we provide an on-demand version, our focus is really on the large enterprise, and what we hear from customers is that they'd rather have it on-premise. As a result we don't compete for the same customers."
Instead, Intalio aims to win market share from Microsoft Dynamics, he said, and perhaps also SAP and Oracle. Perhaps surprisingly, Ghalimi singled out Microsoft as a primary competitor once its Azure platform ships, although he said he expected Oracle will use its Sun assets to bring out a similar offering. "Our only competition today is Microsoft with Azure," he told me. "The Azure platform will definitely be very significant competition to us. They're spending massive amounts of money evangelizing the market."
What do you think? Does Intalio's hybrid cloud-on-premise model set a new pattern for the future or is it a dangerous diversion from the on-demand roadmap?