Centralized vs. distributed computing is one of the age-old pendulum swings of IT. Every few years, the smaller incremental acquisition costs and greater availability of local computing power cause a migration to distributed computing. Then, the lack of manageability and accountability when computing resources are under users’ direct control cause a swing back to a centralized paradigm.
This introductory article and future articles will examine consolidation as a have-your-cake-and-eat-it-too approach to combining the best features of distributed and centralized computing. This article will provide a high-level, abstract view of IT resource consolidation. A future article will examine Linux workloads in virtualized environments on both Systems z and x86 platforms. Next, we’ll look at less-than-full-instance virtualization, consolidation from a Disaster Recovery (DR) perspective, and other topics.
Centralization vs. Decentralization
Let’s put two stakes in the ground: At the fully centralized end, we have a traditional glass house mainframe, and at the fully distributed end, we find a network of PCs; in this day and age, they’re probably running Windows.
With the fully centralized environment, user interaction with the system is limited. Green-screen terminals are the predominant input paradigm. Access to physical system resources is tightly controlled. Operators and systems programmers are the only people allowed to enter the raised-floor data center where the mainframe lives. All devices—DASD, tapes, channel-attached printers—are behind the glass wall. If you want output, you print something and wait for an operator to put it in your hopper. The mainframe may not even have any IP interfaces.
In the decentralized environment, all files exist on storage local to users’ workstations, and files are shared between systems using a peer-to-peer workgroup model (or simply by mailing copies). Printers are either directly connected to desktop machines or are shared between small groups of users, usually via some sort of IP-based protocol, whether Windows SMB printing, lp, or Internet Printing Protocol (IPP). Users have full administrative rights over their own computers and access to network resources isn’t restricted.
Each of these models has advantages and disadvantages. Further, few actual enterprise networks are at either of these extremes; almost all systems are somewhere in the middle.
Advantages of Centralization
Centralization has many benefits. Resources can be centrally managed; important tasks may be given more resources than less-important ones. System administrators know how much capacity is unused at any time and find it relatively easy to forecast growth and future acquisitions.
The difficulty of installing new software packages leads to an environment that’s hostile to malware and the isolation of the system from external information sources means employees are unlikely to be using the system for non-work-related activities. Centralization of user activity makes it much easier to have oversight of projects hosted on the machine than of projects operating in a distributed environment.