If you are currently running HIS 2004 or HIS 2006 in a virtual environment or want to investigate the possibility of doing so, there is one potential performance issue that you need to be aware of if  you are using (or planning to use) the IP-DLC link service included with HIS 2004 and HIS 2006 to integrate with IBM Enterprise Extender on your IBM systems.

You may find that performance may suffer when using HIS and IP-DLC connections in a virtual environment if the applications being used require long running transactions and use large packet sizes. Unfortunately, I can’t provide any specifics around what defines a long transaction time and large packet sizes as it varies across environments. However, we do know that customers that use HIS 2004 or HIS 2006 in virtual environments where 3270 sessions are mainly used don’t seem to see any differences in performance when compared to running HIS on a real (non-virtual) hardware platform.

In instances where the performance was slower in a virtual environment, the applications that were in use were APPC (LU 6.2) applications that were moving a lot of data between the IBM mainframe and the HIS servers.  

I’m sure that the question that comes to mind is around why would the use of IP-DLC connections with HIS 2004 or HIS 2006 result in possible performance issues.

The answer to this question lies in understanding the High Performance Routing (HPR) extension to IBM’s APPN architecture. The intent of HPR was to improve APPN data routing performance and reliability. HPR uses an algorithm called Adaptive Rate-Based (ARB) flow/congestion control to determine the rate at which data will be sent across the APPN network. It does this in order to prevent flooding the network with data. You can read all about HPR (Chapter 8) and ARB (Chapter 9) in the following IBM document:

Inside APPN and HPR – The Essential Guide to New SNA

ARB measures the time that it takes packets to go across the network. A subset of the packets (ARB segments) act as checkpoints containing the time in microseconds since the last checkpoint was sent from the point of view of the sender. The receiver can then compare that time with the time that it measures to determine if the network is slowing down. If ARB determines that the network is slowing down, it will reduce the rate at which is sends data. If ARB determines that the network congestion is clearing, it will increase the rate at which it sends data. This all makes sense assuming that the delays that ARB detects are related to network congestion.

As you can see the ARB algorithm is very sensitive to timing related issues. If there are timing issues on the system that is sending or receiving the data over the HPR connection, this can cause the ARB algorithm to reduce the rate that data is being sent. This is exactly what can occur when systems using APPN and HPR are running in a virtual environment. There can be delays as the data is passed to a virtual machine (VM) because of timing issues related to running in virtual environments.

The following is a link to an IBM FAQ that mentions this issue for HPR on a CS/Linux environment:


The following is a link to an IBM readme that describes the same issue related to running IBM Communications Server in a virtual environment:


As you can see, the potential performance problem can occur with any software that utilizes APPN and HPR to integrate with IBM Enterprise Extender when running in a virtual environment.

If you want to use HIS 2004 or HIS 2006 in a virtual environment using IP-DLC, you should make sure to thoroughly test all of the applications that will be used to make sure that the performance is comparable to that seen when using physical hardware.