I would like to understand better the performance data from ISAM.

The usual answer is typically "it depends", which is true. To circumvent this issue, we can fix some variables to get more precise results, and get the data we need. For example, if it depends heavily on the amount of memory, we will fix it to 8GB, or any other value, so it won't be a "dependency" problem anymore. If it depends on the configuration, we can also fix it as the most basic, 1 reverse proxy, 2 junction, local ldap, and so on.

Bottom line, how many requests per second can ISAM handle? Surely IBM has models to determine the number of ISAM appliances an organization needs to support their workload. It would be nice if we had some type of rule of thumb to determine this.

Surely we can use tools to evaluate this, but the idea is getting an estimate.

There are other areas of concern regarding performance, that I will post in another thread that deal with impact of some configuration options on the performance of ISAM.

------------------------------
Joao Goncalves
Pyxis, Lda.
Sintra
+351 91 721 4994
------------------------------

Hi Joao,

ISAM performance and capacity would depend on the deployed functionality. In case of reverse proxy, the memory requirement is influenced by throughput requirement, session caches, ssl session cache , number of worker threads etc etc. You can refer to the performance tuning guide  to get more finer details on the memory utilization footprints. It is always important to run performance tests mimicking production workload to size the deployment.

For a generic rule of thumb, you can consider 4GB of memory requirement per CPU core available to the appliance.  

The GEN2 hardware appliance is capable of supporting 2400 logins/sec and/or 42000 page access requests/sec. Specifications for Hardware Appliance can be found here .

Do let us know your concerns. At a major part, reverse proxy performance is impacted by the junction servers. A slow responding junction server could held up the worker threads and starve the reverse proxy.  Monitoring is equally important to identify the bottleneck and tune the involved systems. You can find more details on monitoring options here.

Thanks,

Soum

------------------------------
Soumyajit Roy
------------------------------

Original Message:
Sent: Fri October 23, 2020 05:59 AM
From: Joao Goncalves
Subject: Capacity Planning

I would like to understand better the performance data from ISAM.

The usual answer is typically "it depends", which is true. To circumvent this issue, we can fix some variables to get more precise results, and get the data we need. For example, if it depends heavily on the amount of memory, we will fix it to 8GB, or any other value, so it won't be a "dependency" problem anymore. If it depends on the configuration, we can also fix it as the most basic, 1 reverse proxy, 2 junction, local ldap, and so on.

Bottom line, how many requests per second can ISAM handle? Surely IBM has models to determine the number of ISAM appliances an organization needs to support their workload. It would be nice if we had some type of rule of thumb to determine this.

Surely we can use tools to evaluate this, but the idea is getting an estimate.

There are other areas of concern regarding performance, that I will post in another thread that deal with impact of some configuration options on the performance of ISAM.

------------------------------
Joao Goncalves
Pyxis, Lda.
Sintra
+351 91 721 4994
------------------------------