Fake clouds: The last gasp of client-server

You call that a cloud? I do not think that word means what you think it means

Recently, I was pitched SAP HANA as a cloud solution. I checked out SAP's site, and sure enough, there was a multitenanted cloud version of SAP HANA -- a free one for developers, that is. For any serious work, you needed a "dedicated instance" at $539 per month for the Starter edition. The Standard edition costs $3,932 per month.

Oracle, which you've heard a million times by now is all-in with the cloud, provides another example. To get Oracle Database as a Service, the price per OCPU (Oracle Compute Unit) per month starts at $600 for the Standard Edition and climbs all the way to $5,000 for the Extreme Performance Enterprise Edition. Click the Buy Now button and up pops an 800 sales number for you to call.

Did I mention that to get even that far, you need to register for an Oracle account?

The cloud is all about self-service. You go to Salesforce or Amazon Web Services, enter a credit card number, and start working. You may or may not want to develop a lasting relationship with that cloud provider. You pay for what you use and are free to turn off the spigot at any time.

That on-demand model is the future. Yes, there are clear advantages to picking a single platform; ask Netflix about the elaborate microservices ecosystem it has developed on Amazon Web Services to power its Web applications. But you want fluidity, especially when you're starting out. Plus, the hottest area of enterprise tech is container technology that may soon enable customers to pick up and move multipart applications among clouds.

The client-server era depended on long-running, deeply entangled relationships between vendors and IT management and even C-level business management. Back when enterprise tech progressed at a glacial pace, you could only hope a big vendor would eventually, when it got around to it, incorporate substantial improvements.

Our current blistering pace of change simply doesn't support that model. Today, someone develops a great new open source solution, posts it to GitHub, and developers everywhere take it for a spin. If it gets traction, you see cloud providers start offering it as a service. That's been true of Hadoop and its ecosystem along with -- more recently and with fierce velocity -- Docker and its ecosystem. 

It's hard for companies that maintain a client-server legacy to create open cloud platforms without a priori lock-in. Worse are service providers that talk the cloud talk, then actually offer managed hosting, with dedicated hardware and licensed software. The lock-in is total and the expense is high -- and if it isn't high, you have to wonder about the sustainability of the model. Remember application service providers? They proved that one-off approach wasn't scalable more than a decade ago.

There's certainly a role for consultancies in the cloud, in part because customers may need to string together services from multiple clouds, which can be a complicated endeavor. But the cloud services themselves need to be open, self-service, pay-as-you-go, and multitenanted (unless government regulations or performance requirements prevent the latter). Otherwise, "cloud" doesn't mean a thing, and we'll run afoul of the same lock-in and foot-dragging that enterprise computing was notorious for years ago.


Copyright © 2014 IDG Communications, Inc.