IPA on UNIX - Adding LPA nodes to the Landmark Grid

Frankie Hennessey with BayCare Healthcare system here. Does anyone have any experience adding LPA nodes to the Landmark Grid for a UNIX environment?

We are hoping to be able to enhance the throughput of our IPA system so that we have the LPA nodes associated to specific IPA processes and services. Right now we have 1 LPA node with 10 threads we are hoping to increase that to 4 or 6 total LPA nodes on our UNIX environment.

Find more posts tagged with

Process Automation (IPA) - General Discussions

Comments

russel-cornelson

Good morning Frankie,

What version of Landmark Environment are you on?

frankie-hennesseyiv

Hi Russel,

We are on 10.1.1.48.

russel-cornelson

Frankie, over the past few months I have been compiling data from people as they upgrade Landmark to 10.1.1.48 or higher and while the first couple people to report in indicated they required 2 nodes. All other clients reporting in on UNIX indicated they also experienced improvement running 1 node with no restriction on the number of workuntis running simultaneously.

Adding another node is fairly easily done but each node consumes resources; so in my opinion, testing should be done to determine whether or not you need more than 1 node.

The easiest way to test this, in my opinion is to create a process flow:

Start Node(with 10 variables, var1,var2,var3...) > BeginForLoop(10000 iterations) > Assign node (assign each variable a value) > EndForLoop > End Node

Then create a 2nd flow that allows you to trigger this flow X times.

Start Node > ForLoop (X iterations) > trigger node to trigger above flow > EndForLoop > End Node

---------------------------------------

Then to start things off run the flow 1 single time with nothing else running and record the execution time. This is the processing time you are aiming for. So pretend a single run executes in 10 seconds.

Then run the 2nd flow so that X is higher than the max number of flows you can run in a single node. In your case currently 10; so set X to 15. Then analyze the execution time of 10 flows running at the same time.

----------------------------------------

If the flows still process, with an execution time that is "close" to the time of a single run. Then you can increase the number of workunits running in a single node instead of launching a 2nd node. You should repeat this test until you have the system running at the number of simultaneous workunits you are aiming for.

NOTE: While testing in this manner, you should be monitoring your CPU usage to ensure you aren't pegging the Landmark Server CPU. 10.1.1.48 running with the js.compile=-1 setting enabled have seen some dramatic improvements in performance and I now believe it's quite possible to configure the system to overrun the CPU of that server and cause other issues.

frankie-hennesseyiv

Thank you very much for the info Russell. I will be going through these steps with our sys admins to determine our next steps then will get back to you what we find.

frankie-hennesseyiv

Russell,

The following are our results from this test. We are using the one LPA node with 10 threads for both TEST and PROD. It seems that something is not configured correctly?

Single flow running in TEST (Completed in about 2 1/2 min);

Single flow running in PROD (Completed in about 2 1/2 min);

10 flows running in TEST (Completed in about 6 min)

10 flows running in PROD (It never completed after 32 min)

5 flows running in TEST after the 10 completed (Completed in about 4 min)

5 flows running in PROD (I had to cancel since the original 10 in PROD never completed)

Legacy Contributor

just wanted to confirm that the js.compile=1 only for version 10.1.1.48 and higher?

Copyright © 2025 Infor. All rights reserved.