In replacing infrastructure servers, we can choose to:
. Continue to use a distributed design but use Linux servers.
. Consolidate onto fewer larger platforms.
. Replace servers with network- attached storage.
.Combine all three strategies.
Many Small Linux Boxes
The problems of managing many small computer servers for and print serving are well understood, which is what generates interest in the alternatives. The advantage of this is that we can do a piecemeal upgrade in place, as opposed to trying to buy all new systems. Since Linux systems cost little and will probably outperform the software they are replacing, we get better performance at almost no cost.
Consolidation
Server consolidation is very likely to be a piece of the solution, because it puts configuration can be attractive, but unfortunately organizational boundaries and geography often prevent us from getting the full benefits because some small servers just need to be in remote locations or outlying divisions.
Also, vendors price consolidation to reap much of the benefit for themselves; for example, a Microsoft enterprise-class operating system that can replace three servers costs about three times as much as the standard server, and larger hardware is often more costly than smaller. Finally, consolidation is popular with vendors because it generally means buying all new hardware, when much of the older equipment is a sunk cost and may be quite serviceable. Typical organizations should see fairly constant file and print use unless they are expanding employment.
Network-Attached Storage
Network-attached storage (NAS) offers management and cost advantages, although once again actual geography and the need to replace hardware that is a sunk cost can eat up the advantages.
4.2.2 Manage Use of Windows Proprietary Features
The goal here is for those organizations that are not able to make a large-scale move from Windows today to prepare the ground for a move away from Windows to open source when the time is right. This could be in response to a business change, license upgrade, or just when systems are ready for replacement.
We will postpone new licensing as far as possible. We will postpone further use of Active Directory, and look at open alternatives for directory and security. This has the benefit of potentially reducing licensing fees while making integration with non-Windows systems simpler.
We will avoid upgrades in general, as they are only justified by new features. We will stay on NT4 rather than Win 2K, and on Exchange 5.5 rather than 2000. We will prepare the ground for open source types of support by looking at community support options over, for example, MS Premier Support.
Systems management tools should be built using open source development tools such as Python, not platform-specific APLs, and tools that are purchased should be open source or cross-platform if possible, or else inexpensive if platform specific. Systems management in particular is an area where good people inside good companies often make large and well-designed investments, only to find themselves trapped within proprietary code.