Measuring Cloverleaf Performance

I am in the process of upgrading from CL 5.6.2 to CL 5.8 and want to be certain that the performance on 5.8 (with new hardware and O/S (and probably with vmware)) will be adequate.

The general approach is to compare Cloverleaf generated thread statistics on current the Prod host with thread stats on the new host after processing an identical message load over the same time interval.

I have a script that zeroes out the thread stats in all of the Prod sites and another that gathers thread stats at various intervals.

There’s also a tcl script that runs on a separate host that replays SMAT messages from between user supplied start and end times and observes the inter-message time intervals for sending. Lastly, there’s an “ack farm” on another host that receives all the CL outbound messages and generates an HL7 ack. Outbound threads in the NetConfigs are modified to send  to the ack farm.

I am a bit confused by some of Cloverleaf’s latency measurements.

Here are a few examples from a two hour prod sample:

Are any of these Latency measurements useful for comparative performance purposes?

IB Latency    : 0.000

OB Latency    : 13.497

Total Latency : 16.625

IB Latency    : 0.000

OB Latency    : 0.023

Total Latency : 17.293

IB Latency    : 10.895

OB Latency    : 1.199

Total Latency : 1.201

======

Are Time on Queue (T on Q) or Xlate time (X time) good measurements for comparing performance?

Inter-thread Statistics

Sent    Recvd   pxqd X time  T on Q  Latency Thread name

   620       0    0   5.307   0.000   0.000 AZ_RadOrd_RIMS                  

     0     620    0   0.000   0.568   1.201 AZ_Mill_RadOrd  

TIA for any comments or suggestions.