|Modern Middle Manager
Primarily my musings on the practical application of technology and management principles at a financial services company.
The New IT Crisis
Saturday, December 21, 2002 This article from Marc Andreessen discusses what he believes is the next wave of IT innovation -- driving down costs by making computing resources as automatic and easy as getting a dialtone when you pick up the phone. I sincerely hope we get there. IT should be spending more time on enabling business processes and adding value than spending resources on maintenance.
There are several obstacles to implementing this kind of "utility computing." #1: How to you propose to create a new server on demand? How will it be configured with the right settings and custom software? There is a lot of talk about pushing OS images to spare server appliances on demand, installable software modules, automated patching and software robots that perform monitoring and self-healing. It's a great concept and one I'd love to see in practice. I think it will be a bit before we get there.
What practical steps can be taken NOW to drive down maintenance time and expenses? Is it possible to achieve SOME of these objectives today? Let's examine some of these issues:
1. Creating a new server on demand. As an alternative to SysPrep, we are using virtual servers which are actually a number of files on a hard disk, loaded and managed by VMWare's GSX Server. By creating a basic set of server images (such as a Windows 2000 member server, perhaps one with IIS or similar software -- this doesn't work with Exchange or SQL Server), it's possible to rapidly deploy a server by copying those virtual server files to another system, start it up without networking, rename it, turn networking back on, put it in the domain and get it functioning. Not quite the same as pushing a fully-functional server image but it's much faster than simply reinstalling Windows 2K or even SysPrep. For Linux servers, the process is even quicker -- rename, readdress, be done with it.
2. Installable software modules and automated patches. We've tried Microsoft's SMS (harken back to In Living Color...hated it!) and LANdesk Software's LANdesk Manager deployments (hated it!). We may evaluate something closer to the installable software module idea with InstallShield's AdminStudio. Updates to Microsoft servers and desktops are currently done using St. Bernard's UpdateExpert program. Really, who wants to look for a software package that requires hours of specialization from your staff and sets itself up intrusively on every [Microsoft] server? Isn't there another way to push updates and/or full software packages? And Linux is not leading the way here, although we're using the Debian version so at least we could schedule an apt-get update to upgrade the systems. It's still pulling software as opposed to pushing it out, though, limiting the IT department's control.
3. Software robots. There is software now that attempts to do self-healing (SMS and LANdesk come to mind again) although I have been unimpressed by LANdesk's ability -- in fact, it's self-healing module broke Microsoft Outlook's ability to open attachments and interfered with some of our scanning software. I think the technology still needs work. Using robots to monitor processes in [near] real-time does have some promise, though, and can be done now. I think of the article I read on General Electric's initiative to create a real-time enterprise. We have created a scorecard for several critical processes that we monitor on a near real-time basis as well. While not as comprehensive or currently as pretty as GE's, it has proven to deliver bad news to quickly enough to resolve problems before an end-users and clients notify us of service level issues.
I eagerly anticipate the day when computing resources are like Legos and their maintenance is near-automatic. I sincerely hope that Operating System and Server vendors are moving in that direction.
posted by Henry Jenkins | 12/21/2002 02:08:00 PM
Comments: Post a Comment