
A Principled Technologies sizing guide 4
Red Hat Enterprise Virtualization sizing guide
Based on our testing, we developed the sizing guidelines in Figure 1.
Number of guests per physical core
Disk activity (IOPS) per guest
Network traffic in Megabits per
second per guest
Figure 1. Sizing requirements for two types of users.
USING LOGIN VSI TO LEARN ABOUT A SERVER’S CAPACITY
To determine our server’s virtual desktop capacity, we used the Login
Consultants Virtual Session Indexer 3.5 Release 1 benchmark. Login VSI consists of
several workloads that perform a range of tasks to simulate different kinds of users. The
results show the maximum number of virtual desktops a server can support for a
maximum response time on a particular set of tasks. (For more detailed discussion of
how Login VSI works, see Appendix A.) In addition to the performance scores from the
Login VSI benchmark, we measured server and client network bandwidth, and server
CPU utilization, memory usage, and disk activity.
After all desktops are idle, Login VSI incrementally logs users into virtual desktop
sessions and begins workloads on each. Login VSI measures the aggregate time it takes
to complete seven typical office operations from each session, and reports it as
“response time”. As the number of guests running increases, response time also
increases. After a certain point, a long response translates to unacceptably slow
performance for end users.
Figure 2 shows the average Login VSI response time for all active sessions we
recorded during our office worker test. We ran 110 users and the server was able to
support this load without reaching what Login VSI considers an unacceptable response
time. Figure 2 also shows the approximate user counts where the processor utilization
was at 40% and 80%. Generally, 40% utilization is very comfortable, but once a server
hits 80%, it is getting close to saturation and leaves little room for growth or changing
user needs.
Komentáře k této Příručce