Tc/foip – Kofax Communication Server 9.1.1 User Manual
Page 31
Environment Guide
Version 3.00.04
© Copyright Kofax, Inc. All information is subject to change without notice.
31
(3) This indicator stands for both local and remote disk latencies if Tandem Server is being used.
Mapping among key performance indicators and corresponding Windows performance counters
(1) The remote disk write counters include a sum of as network latency between primary and secondary
master and the disk access time itself.
Average counters (“Avg.”) indicate the average of the particular value between two subsequent
performance monitor measures.
Peak counters (“Peak”) indicate the peak value between two subsequent performance monitor measures.
After the deployment has been already done, it is necessary to continue observing the actually used
resources of the particular TCOSS Server configuration and adapt the resources allocation if necessary.
Please refer to the chapter Virtualized Operation.
TC/FoIP
KCS FoIP solution is based on the third-party stack called XCAPI. According to its vendor, there is a
requirement to allocate one or two processor kernels exclusively for the VM where the XCAPI is running.
This should be accomplished by setting CPU affinity of own VM to particular processor kernels (e.g. 0 and
1), while setting CPU affinity of ALL OTHER VMs on the same ESX server to remaining processor kernels
(e.g. 2 and 3, in the case of 4 processor kernels on the ESC server).
However, according the VMware
literature this solution has significant drawbacks (see “Potential Issues
with Affinity” i
. The most important seems
to be the fact, that the ESX server is not able to guarantee CPU reservations for those VMs which set
manual CPU affinity in this way.
Therefore following approach has been chosen to support FoIP solutions under ESX server:
1.
No manual CPU affinity is required
2.
Minimum CPU bandwidth and RAM must be granted for the VM running TCOSS and FoIP (based on
the following table) in order to increase the probability of fault-free operation
3.
In the case of any problems (like interrupted faxes etc.) the customer will have to reproduce the
problem on the physical hardware in order to get support from Kofax
Used abbreviations
Explanation
Performance Object
Performance Counter
Instance
Name
RTT
ls1
RTT
FoIP
Round-trip-time between
TCOSS and LS1
servers/TC/FoIP Servers
TCOSS Links
Avg. Packet-Ack Time ms
Peak Packet-Ack Time ms
L.XY
RTT
prim-sec
Round-trip-time between
primary and secondary
Master (that occurs during
the remote-disk access)
TCOSS Links
Avg. remote Disk Network
Delay ms
Peak remote Disk Network
Delay ms
None
RTT
storage-media
Round-trip-time between
Storage and Media
Servers (in the case of
ASP system)
TCOSS Links
Avg. Round Trip Time ms
Peak Round Trip Time ms
0, 1, 2, or
3
Average Write Disk
Latency [ms] ]
(1)
Average duration of a
write disk access in ms
TCOSS Disk
Avg. local Disk ms/Write
Avg. remote Disk ms/Write
None
Peak Write Disk
Latency [ms]
(1)
Peak duration of a write
disk access in ms
TCOSS Disk
Peak local Disk ms/Write
Peak remote Disk ms/Write
None