Page 132 -
P. 132
6: A Most-Significant Step—“Virtualizing” Your IT Systems 97
If thin-client computing and client virtualization in general have all those
significant benefits, then why aren’t all desktops thin clients in the work-
place? It’s about personalization of the client, since that’s been the norm for
many years. However, IT vendors are in the process of changing that. The
IBM Virtual Client Solution, for example, can take a full-featured desktop
OS such as Windows XP and run it on a virtual server. Under this new
approach to thin-client computing, users can retain all the PC experience lost
with simple graphical user interface (GUI) virtualization, and the IT team
can gain the savings brought about through centralized PC management.
A Further Look at Reasons for Creating Virtual Servers
Consider this basic scenario. You’re in charge of procuring additional
server capacity at your company’s data center. You have two identical servers,
each running different Windows applications for your company. The first
server—let’s call it Server A—is lightly used, reaching a peak of only 5 per-
cent of its CPU capacity and using only 5 percent of its internal hard disk.
The second server—let’s call it Server B—uses all of its CPU (averaging 95 ptg
percent CPU utilization) and has basically run out of hard disk capacity.
(That is, the hard disk is 95 percent full.) So, you have a real problem with
Server B. However, if you consider Server A and Server B together, on average
the combined servers use only 50 percent of their CPU capacity and 50 per-
cent of their hard disk capacity. If the two servers were actually virtual servers
on a large physical server, the problem would be immediately solved because
each server could be quickly allocated with the resource each needs. In newer
virtual server technologies—for example, Unix Logical Partitions (LPARS)
with micropartitioning—each virtual server can dynamically (instanta-
neously) increase the number of CPUs available by utilizing the CPUs cur-
rently not in use by other virtual servers on the large physical machine. This
idea is that each virtual server gets the resource required based on the virtual
server’s immediate need.
Figure 6.4 shows typical server utilization for stand-alone servers (that is,
no virtualization). The multimillion dollar mainframes are typically utilized
on a 24/7 basis at least partly because of the large financial investment.
Mainframe “batch” processes such as running daily, weekly, and monthly cor-
porate summary reports are typically CPU-intensive and are run at night and
on the weekends. The small department Windows server (labeled “Intel-
based” in the diagram) is not typically used at night or on the weekends.
Creating virtual servers of those Intel-based servers not only allows much