GM caches out

Economic times may be uncertain, but that hasn’t held back Tony Scott. As chief technology officer at General Motors Corp. and head of the company’s Information Systems & Services group, he is pursuing IT infrastructure initiatives ranging from wireless LANs to content caching.

What infrastructure technologies top your IT agenda right now?

We continue to see pressure in the network, from a bandwidth and utilization perspective. So we’re investing in [content]-caching [servers] this year fairly heavily. We’re also investing in security fairly heavily [by] reimplementing our network architecture into security zones.

What benefits did content-caching edge servers bring?

There were lots of areas in our network that have pressure on them where the cost to upgrade would be fairly significant. [Through] strategic placement in the network, as well as educating the development community to design applications that can take advantage of the caching, we’re seeing pretty nice benefits.

What are the biggest challenges you face in rolling out those initiatives?

The biggest problem is complexity. When you think about all the layers of technology it takes to put out a simple Web application today, the opportunities for misconfiguration or creating a bottleneck unintentionally are just enormous. And . . . even if you get it right initially, there’s always the need for constant changes in the environment, whether it’s upgrading software or scaling issues or any number of different things.

GM has been upgrading its desktop and server infrastructure to Windows 2000 for some time. What makes that process so time-consuming? When you’ve got to migrate tens of thousands of users from one desktop to another and preserve all the applications and settings and the user look and feel, that’s an enormous task and still one that hasn’t been solved, really. There are lots of different tools, and each helps in its own way, but that’s still the single biggest issue.

One of the fundamental things we’ve done that pays off more than anything else is simplification. Before you even start doing the migration, eliminate as much as possible. In every area we found old, obsolete applications that we could simply turn off or make go away as part of the migration, thereby simplifying the rollout.

In January, you were considering using Office XP and .Net Web services to link to server-based data on sales and information on the status of orders. What have you done so far?

Not a lot yet. We’re getting more interested in the security model around Web services, and we’re sorting a bunch of things out there. I guess that’s the one thing. There’s a fair amount of grayness in [Web services security], and that’s holding up any broad deployment at this point.

What are your criteria for prioritizing projects in today’s uncertain economic climate?

Security takes the highest priority security- and safety-related things. In our manufacturing environments, anything that comes up in those areas always floats to the top. The next thing is business demand. We use the business community to help prioritize projects beyond that and they are not shy about telling us what they want, and when and where they want it.

We have a corporate mandate for common and global [projects]. Given a dollar to spend, and one is a regional, local project and the other is a global, common sort of project, the global, common one will tend to float to the top.

What emerging technologies do you find most compelling?

The obvious one is wireless [Ethernet], which we’re beginning to take strong advantage of, particularly in the factory environment. And I’m excited about the new tablet PCs. When you combine the tablet PC with wireless, you create a whole new way of thinking about computing. We’re doing an early look at that.

I’m particularly excited about this Handspring Trio that was just announced, the color one. I think all of that’s quite intriguing.

Which would fit better at GM: a Palm OS-based device or a Windows CE device?

From an infrastructure standpoint, what we have focused on is the middleware piece. We know the end device the PDA is going to go through a very rapid evolutionary cycle, so we’ve tried to concentrate on the middleware that lets us talk to whatever device happens to be there. That’s helped us already. Even in the [past] year or so, we’ve gone through two or three generations of handheld devices, and we have not had to change the application one iota.

What’s your biggest gripe about the IT your teams work with?

One is the pain of upgrades. I don’t see anybody that’s really cracked the code on that. Every opportunity we have [had] to do a software upgrade or a hardware upgrade [has had] a fairly high level of pain.

Quality and complexity probably go hand in hand in terms of No. 2. The number of patches and emergency upgrades and so on still continues to be relatively high. And most of that you can pin on inadequate testing or just the sheer complexity of the product. That complexity factor is huge.

If you could replace one legacy IT infrastructure right now with state-of-the-art technology to make your job easier no strings attached what would you choose?

In the factory environment, we still have a bunch of very old technologies. These are DOS-based applications, for example, probably running on 286-class machines at best. They still work, so there’s no business case for replacing them, and yet the support cost is high. If you could consolidate and centralize a lot of these applications, you could take a fair amount of cost out.

More medium term, we have a lot of first-generation client/server systems … where there’s a thick client on the desktop, and I would replace those with a more Web-based configuration. Those cost a lot in terms of ongoing support and maintenance. Every time we’ve got to go upgrade the desktop, we have to retest hundreds and hundreds of these applications, and that causes a fair amount of support costs.