Hello all, 

i have an inquiry if i'm going to request resources for traditional IBM app connect enterprise to can achieve 10000 RPS - Request per second - , 

What is the best and optimum resources will be required for it noting that there will be no bottlenecks on DB calls or core BE calls average time response is 10 ms per each . 

• how many instances for the bar of the application which will recieve this number of requests ?
• what is the best requirements for the cores , storage and RAM required for this integration node ?
• to handle this enoromus number of calls per second with immediate response would it be better building this app on IBM MQ or rest API will be working fine?

------------------------------
Ahmed Elelaimy
------------------------------

Hello Ahmed,

With this straight forward requirement and the amount of details provided, you will get multiple options and none will work as-is without collecting more details.

Basic strategy could be:

1. We have 1000ms in each second.
2. One thread can serve 1000 / 10 = 100 requests per second.
3. 10 000 rps / 100 req = 100 threads. 

This is the most optimistic scenario when no delays are expected, message is not persisted, not mapped, not logged, etc.

Then it comes to that extra bunch of requirements, specifically to NFRs:

1. Are you ok to lose messages in case of hardware failure?
2. Do you need a full history of messages stored in logs or otherwise tracked on the Integration layer?
3. Are your consumers able to send requests and receive responses on MQ queues, or they prefer some other protocol, e.g. HTTP - ReST, SOAP?
4. Do you need to map messages from one format to another, e.g. JSON to XML and vice versa for responses?
5. Do you have multiple different backbends serving different types of requests, e.g. one backend for orders, another backend for customer info, etc.?
6. What is the size of each request and response? Do you have to parse it (in memory) or you will proxy it as-is to the destination?
7. Do you have a load-balancer that can forward requests to the "right" instance?
8. Any complex scenarios with request or response aggregation?

There are many more questions that will drive the extra capacity you may need to serve all your requests.

Hope this helps as a starting point in your requirement analysis.

------------------------------
Jevgenijs Jefimovs
------------------------------

Original Message:
Sent: Tue February 10, 2026 08:15 AM
From: Ahmed Elelaimy
Subject: environment creation to receive 10000 RPS -Request per Second-

Hello all, 

i have an inquiry if i'm going to request resources for traditional IBM app connect enterprise to can achieve 10000 RPS - Request per second - , 

What is the best and optimum resources will be required for it noting that there will be no bottlenecks on DB calls or core BE calls average time response is 10 ms per each . 

• how many instances for the bar of the application which will recieve this number of requests ?
• what is the best requirements for the cores , storage and RAM required for this integration node ?
• to handle this enoromus number of calls per second with immediate response would it be better building this app on IBM MQ or rest API will be working fine?

------------------------------
Ahmed Elelaimy
------------------------------

Hello Ahmed,

I agree with Jevgenijs. We don't have sufficient information that say anything usefull about it. A good starting point could be looking into the performance report where different scenarios are explained and measured.  Performance Report ACE V13 https://www.ibm.com/support/pages/node/7235050

Depending on your specific scenario you can make some estimates. You will need to consider where your bottleneck will be compute or io operations and the choose to scale vertical or horizontal. Also note that ACE has many built-in metrics you can check see https://www.ibm.com/docs/en/app-connect/13.0.x?topic=performance-resource-statistics 

My experience with sentences like "no bottlenecks on DB calls or core BE calls average time response is 10 ms per each" is that you have to be very careful with those claims. Especially when you are scaling up and are using ACE in a correct way.  

Let me know if you need some additional assistance. 

 

------------------------------
Jef Jansen
Integration Architect
IntegrationDesigners.com
Kontich
------------------------------

Original Message:
Sent: Wed February 11, 2026 06:00 AM
From: Jevgenijs Jefimovs
Subject: environment creation to receive 10000 RPS -Request per Second-

Hello Ahmed,

With this straight forward requirement and the amount of details provided, you will get multiple options and none will work as-is without collecting more details.

Basic strategy could be:

1. We have 1000ms in each second.
2. One thread can serve 1000 / 10 = 100 requests per second.
3. 10 000 rps / 100 req = 100 threads. 

This is the most optimistic scenario when no delays are expected, message is not persisted, not mapped, not logged, etc.

Then it comes to that extra bunch of requirements, specifically to NFRs:

1. Are you ok to lose messages in case of hardware failure?
2. Do you need a full history of messages stored in logs or otherwise tracked on the Integration layer?
3. Are your consumers able to send requests and receive responses on MQ queues, or they prefer some other protocol, e.g. HTTP - ReST, SOAP?
4. Do you need to map messages from one format to another, e.g. JSON to XML and vice versa for responses?
5. Do you have multiple different backbends serving different types of requests, e.g. one backend for orders, another backend for customer info, etc.?
6. What is the size of each request and response? Do you have to parse it (in memory) or you will proxy it as-is to the destination?
7. Do you have a load-balancer that can forward requests to the "right" instance?
8. Any complex scenarios with request or response aggregation?

There are many more questions that will drive the extra capacity you may need to serve all your requests.

Hope this helps as a starting point in your requirement analysis.

------------------------------
Jevgenijs Jefimovs

Original Message:
Sent: Tue February 10, 2026 08:15 AM
From: Ahmed Elelaimy
Subject: environment creation to receive 10000 RPS -Request per Second-

Hello all, 

i have an inquiry if i'm going to request resources for traditional IBM app connect enterprise to can achieve 10000 RPS - Request per second - , 

What is the best and optimum resources will be required for it noting that there will be no bottlenecks on DB calls or core BE calls average time response is 10 ms per each . 

• how many instances for the bar of the application which will recieve this number of requests ?
• what is the best requirements for the cores , storage and RAM required for this integration node ?
• to handle this enoromus number of calls per second with immediate response would it be better building this app on IBM MQ or rest API will be working fine?

------------------------------
Ahmed Elelaimy
------------------------------