This article discusses the effects of running multiple RPM (Realtime Performance Monitoring) probes on Juniper MX routers, particularly the MX80 model. It highlights how the number of RPM probes varies based on platform capacity, resources, and the Junos version in use, while also detailing the impact on router performance, CPU, and memory utilization.

When running a large number of concurrent RPM probes on a Juniper MX router, administrators may notice the following symptoms:

• Increased CPU Utilization: Running multiple RPM tests simultaneously can lead to higher CPU usage, especially when dealing with numerous ping requests.

• Memory Overload: As more probes run, the router consumes more memory to store state information, leading to potential memory resource strain.

• Reduced Performance: A high number of RPM probes may cause overall performance degradation, affecting the router's ability to process other tasks and manage network traffic.

• Probe Failures or Timeouts: Exceeding resource limitations could result in RPM probes failing to complete, causing inaccurate performance data.

The number of RPM probes you can configure on a Juniper MX router depends on platform resources and the version of Junos being used. Below are key points to consider for efficient RPM probe configuration:

1. Platform Capacity: The Juniper MX80 router can support a significant number of RPM probes, but you must account for the available CPU and memory resources. These resources will determine how many concurrent probes can run without impacting overall router performance.
2. Configuration Example: Here is an example configuration to set up an RPM probe on a Juniper MX router:

set services rpm probe PING test PING-TO probe-type icmp-ping
set services rpm probe PING test PING-TO target address 10.229.3.1
set services rpm probe PING test PING-TO probe-count 14
set services rpm probe PING test PING-TO probe-interval 1
set services rpm probe PING test PING-TO test-interval 1
set services rpm probe PING test PING-TO source-address 10.229.3.2

• Performance Impact of 200 Simultaneous RPM Probes:

• • CPU Consumption: Each RPM test sends ICMP packets and waits for responses. With a configuration of 14 pings per test, the router processes 14 ping requests per second per test. Running 200 such tests concurrently will increase CPU utilization as the router manages and processes all requests. During peak periods or high traffic conditions, the router's CPU utilization could spike, potentially affecting performance.

• • Memory Consumption: Each RPM test requires memory to store results, track state information, and manage timeouts or retries. For 200 concurrent tests, memory usage will increase. Each probe might consume a few kilobytes of memory, resulting in an increase of approximately 10-50 MB of memory usage. However, if the router is already running other memory-intensive processes, this could lead to resource contention and performance issues.

By balancing the number of RPM probes with the router's available resources, you can ensure effective monitoring without negatively impacting router performance.

For details on how to configure RPM probes and more information about the supported limits, please refer to Juniper’s official documentation on RPM probes:

Juniper Documentation on RPM Probes