beautypg.com

3 virtualization deployment, 1 cpu, Virtualization deployment – Kofax Communication Server 9.1.1 User Manual

Page 33

background image

Environment Guide

Version 3.00.04

© Copyright Kofax, Inc. All information is subject to change without notice.

33

Resources Estimation and Key Performance Indicators for IPPrinter
Receiving fax channels [x]/
Printouts in fax pages / minute
IPPrinterMetrics

10
20 fax pages / min

20
40 fax pages / min

30
60 fax pages / min

Resource capacity to be planned on the ESX server:
CPU [MHz]

1000

1200

1400

RAM [MByte]

(1)

1024

1024

1024

Average IPPrinter disk kBytes/s

100

125

150

Average IPPrinter disk IOPS/s

22

24

28

Average network kBytes/s

70

90

110

Key performance indicators that must be measured and fulfilled for the proper operation:
none

(1) If more than 1 IPPrinter instance is being used 2048 MByte RAM should be planned

5.3 Virtualization Deployment

This chapter provides a short overview on different possibilities how to deal with system resources on the

ESX host.

5.3.1

CPU

The ESX server handles the CPU resources in terms of “MHz bandwidth” by simply adding the MHz CPU
cycles of all available CPUs together (e.g. having 4 dual-core 3.6 GHz CPUs in the ESX server would give

total CPU bandwidth of 4x2x3600=28800MHz). There are two possibilities how to allocate this CPU

bandwidth to the particular VM:

1)

The shares method (fair share principle)
This method is clearly most effective method for the system administrators, guaranteeing the most
effective utilization of resources on the ESX server. On the other hand, when the administrators add
more and more VMs to the ESX server, this decreases the CPU resources also for the other VMs and
therefore it is not suitable for real-time applications like TCOSS Server, but may be used for other KCS
applications like IPPrinter.

2)

Explicitly grant minimum CPU bandwidth (in MHz)
The particular VM always receives the minimum amount of CPU cycles despite of the fact whether it
needs them or not

– therefore this method is not being favored by the VMware, but it is a clear

precondition for the VM running real-time application. If the necessary resources could not be granted,
the ESX server simply does not start particular VM. Furthermore, due to the Timekeeping issue in the
virtualized environment (see

http://www.vmware.com/pdf/vmware_timekeeping.pdf

) it is absolutely

necessary for the messaging server to receive enough CPU cycles in order to keep its time
synchronized with the real-time (especially important for the messaging servers due to legacy
reasons).
Note that this method is required for KCS real-time applications (TCOSS Server).

The CPU for the particular VM may be allocated by the context menu “Edit Settings” via the VI client: