Kofax Communication Server 9.2.0 User Manual

Page 36

Environment Guide

Version 9.02.00

One of the main reasons for having two distinct RAM setting for the VM is the swap file size estimation for

the particular VM, consider examples:

If configured memory (Hardware panel) equals to 512 MByte, but reserved memory equals 0, then the

size of the swap file would be 512 MByte

If configured memory (Hardware panel) equals to 512 MByte and reserved memory equals to 256

MByte, then the size of the swap file would be 512MByte

– 256MByte = 256 MByte as the ESX server

would potentially need to swap only 256 MByte of allocated RAM as 256 MByte is granted to be always

available for the VM

If configured memory (Hardware panel) equals to 512 MByte and reserved memory equals to 0

MByte, then the size of the swap file would be 512MByte

– 512MByte = 0 MByte as the ESX server

wouldn‟t need to swap anything as the whole RAM is granted for the VM.

Note that granting enough memory is required for KCS real-time applications (TCOSS Server).

5.3.3

Disk

It is not possible to explicitly allocate minimum disk bandwidth for the VM under the ESX server.

Furthermore, there will be typically no local hard disks connected to the ESX server. Instead, one of the

network storage systems will be used (SAN, NFS, NAS).

Typically, network storage arrays (SANs) are divided into logical units (LUN), and each LUN may be

partitioned into several partitions. The ESX server uses following storage addressing scheme:

:::[] (disk partition is optional)

For example, vmhba1:0:0:1, which means host bus adapter 1, SCSI target 0, LUN=0, partition 1:

VMware uses its own file system called VMFS where all VMs are stored in terms of .vmdk files.

Per LUN/partition only one VMFS volume is allowed, but on the other hand, one VMFS volume may span

over several LUNs/partitions. Administrators may use fewer, larger VMFS volumes, but also more but

smaller VMFS volumes.

One advantage of using smaller VMFS volumes is higher throughput for the particular VM due to less

contention on each VMFS volume due to locking and SCSI reservation issues. Also the disk shared

reservation policy works per LUNs, so it seems to be reasonable to request a dedicated LUN with a VMFS

volume for TCOSS file structure

(VMware recommends in the storage best practices guide, that heavily used VMs do not access the same

VMFS concurrently and that they are spread across multiple VMFS volumes).

Consider also the possibility to use the shares method (similar as with the CPU) to set the relative priority

with respect to the other VMs to get the disk bandwidth: