We’re continuing to see more developments in the cloud space that go beyond the obvious (the hype) and address some of the important issues. One in particular is also demonstrating some important facts about the cloud and cloud services; it’s the “Database.com” offering from Salesforce.
Cloud databases have been an issue of increasing importance because they’re essential for the cloudsourcing of any team or company application and because they represent a new dimension in security risk for enterprises. Amazon’s EBS was the proximate cause of that company’s recent cloud outage, so cloud databases also demonstrate the new dimension in vulnerability that this sort of distributed technology can bring. Enterprises need a way of harmonizing cloud use with data security or they’re not going to the cloud with anything that’s important, and that would relegate the cloud to hosting websites or testing/piloting applications.
Database.com is first an attempt to integrate strong security into a cloud DBMS (an RDBMS to be specific). It includes strong authentication at the API call level, meaning that every access attempt is verified, and by-row tabular security rights within the DBMS. All of this is good stuff, and for many enterprise applications it will help relieve security fears. But it’s not enough by itself.
No matter what any vendor says, mission-critical enterprise data isn’t likely to go into the cloud; the career risks for anyone making that decision are profound according to the results of our spring survey. None of the enterprises we asked said they believed they would cloudsource a mission-critical DBMS. So it would appear we’re at an impasse, right?
Not so fast. Database.com is also an example of a database model, DBMS-as-a-service. It’s always been possible to visualize “data” in multiple ways; as disks, as file systems, as DBMSs. That multiplicity of vision translates into a multiplicity of models. You can send disk commands to a database, or file-system commands, or you can send DBMS queries—SQL, for example. When you send the low-level commands, you drag disk I/O over a connection to the cloud if you cloudsource the data, and when you send high-level commands you receive the results of a query and not the ten million records you might have spun through to get those results.
OK, fine, but this is still DBMS-in-the-cloud. It is, unless you turn the tables. Instead of looking at the DBMSaaS as something the cloud offers, how about if the enterprise offers it? Suppose that in a hybrid cloud, the DBMSaaS queries were made by the cloud applications back into your data center? That model is readily supported by modern back-end repository strategies and DBMS appliances. With tabular joining in an RDBMS, it would be possible to create a database that was partly stored in the cloud but whose sensitive elements were back in the enterprise.
Even this may be a model of an even deeper and more important issue. Enterprises say that basic platforms (IaaS) in the cloud are, at reasonable levels of utilization and with reasonable availability enhancements, about 75% more costly than internal servers. That says that the basic business model for IaaS can’t be successful in securing wide penetration of cloud computing into mission-critical apps even if you solve security and availability concerns. But services can be offered from cloud infrastructure, and efficiency in both the resources needed for the service and the way the service can be linked to enterprise computing/business activity can be more compelling. Outsource firm Virtela, who has already created an interesting umbrella VPN service as a kind of VNO across multiple operators, is also launching a cloud service set based on the same framework. The idea is to take applications like security in the mobile space or application acceleration and make them into “services” of the cloud. These are more easily introduced than competing architectures for mission-critical apps, and enterprises in our survey seem to think that sort of thing is the right way to go.
So do carriers, of course. Verizon is clearly looking at this same model, as well as BT. KT, while making some waves by promising IaaS services that are more cost-effective than Amazon’s, is also planning higher-level cloud-based services, and we’re told that they believe there’s more money in the services space than in basic IaaS.
All of this, of course, gets us back to the notion of “SOA clouds” and the need to think of applications as being cloud-optimized. The SOA architecture facilitates the consumption of application components in service form, delivered either through RESTful interfaces or more rigorous SOAP connections (which is how Database.com works, by the way). Microsoft and IBM have both been working with their customers to move thinking in this direction, and the results are becoming clear by the number of enterprises who now think more in SOA terms than in terms of virtualization for their clouds.