I enjoyed dinner a few nights ago with the CTO of a privately held direct/retail sales company whose name you’d most certainly recognize if I had permission to give it to you. Suffice it to say, his company sells a brand-recognizable product you might find in just about any kitchen anywhere in the world today. This fellow was brought on board to helm an IT organization whose budget amounted to nearly 6 percent of overall revenues for the global company. He got the job probably in no short measure because he told everyone they were spending way too much on business technology. After a couple of years, he fulfilled his promise to whittle down IT expense to only about 2 percent, and he wanted to tell me how he did it.

I warn you in advance, this isn’t one of those “we moved everything onto the mainframe again” stories. In fact, one of the steps he took—and his voice and mannerisms suggested it wasn’t a step he has the same confidence about as some of his other cost-cutting measures—was to get rid of the mainframe and to rehost his Enterprise Resource Planning (ERP) on Oracle and Sun gear. More on this later.

He explained that he looked hard at server virtualization as a consolidation technique for the hundreds of servers in his data center, but took a pass on the technology. He was relieved when he discovered that I wasn’t an x86 hypervisor computing advocate, because, based on his testing, neither was he. 

He found that hypervisors such as VMware yielded little in the way of their touted resource utilization efficiency improvements that couldn’t be achieved in other ways. For example, his Linux servers were already doing about the best they could with their processors, memory, and storage—without a hypervisor. 

Basically, he adopted a different consolidation philosophy that can be summarized simply: If an application could be provided any other way than by his shop—say, using an external service provider or a cloud-based service bureau—and if it wasn’t contributing to the core activity of the business, he outsourced it. ERP was kept internal, as were new applications that were being developed pointedly to facilitate business initiatives. Email, however, was gone. Logistics, gone.  Most user productivity apps, outsourced. 

This approach, which I’m increasingly hearing in different variations from many companies, flies in the face of the prior meme of IT. Just a few years ago, the dominant mantra was very different: If it isn’t bolted down, automate it. Very little analysis was performed to weigh or measure the comparative benefits of an automated process vs. a manual one, whether in terms of hard numbers (the cost of infrastructure and labor to support automation) or soft factors (customer satisfaction). We were entering into the “there’s an app for that” phase, and everyone was on board for the ride.

Bottom line: The current economy does seem to be having an impact on computing memes by moving us back to a model in which IT cost is balanced against IT contributions to the business.  “Automate everything” is giving way to a more sensible analytical process: cost-benefit analysis framed in terms of business activity. That’s a good thing.

The not so good thing is that this fellow incorrectly perceived the mainframe as a cost center that couldn’t be justified in the face of a less expensive Oracle Real Application Clusters (RAC) option running on x86 hardware. At least, that’s what he thought. The performance of his Oracle ERP following migration off the Big Iron was poor. He had his team rewrite Oracle’s own code to fix certain inefficiencies, then turned the results over to Oracle.  So far, they’ve ignored his contributions. 

I noted that banks had created exit mods in the ’90s to improve the fit of mainframe software with their business processing requirements—changes that were embraced by IBM only after they had gone viral among many banks. Perhaps his “mods” to Oracle could go mainstream in the same manner? He worried, as did I, that such improvements are now falling prey to the there’s an app for that ideology embraced by consumer computing organizations, of which Oracle is arguably a member. The fixes his team developed may be lost in an ever-growing “yellow pages” of user-submitted apps.

Moreover, I observed that he may want to rethink the transition he made away from the mainframe, noting the many benefits of the latest Big Iron rig, the z114—including competitive pricing with x86 servers, a Linux operating system, and most of the chewy goodness that goes along with mainframe computing. His eyes lit up at the prospect. I’m staying tuned to see what happens. These are interesting times.