Network Monitoring Speed Tests
- How a speed test works
- What are the speed test settings
What you are going to learn:
The speed test feature is available between any monitoring agent inside the organization, including public monitoring agents. A Network Monitoring Session is not required to launch a speed test between two agents.
When creating a speed test, two agents must be specified. Automatically, one agent will be selected as the client (initiating the connection) and one agent will be selected as the server (listening for the incoming connection). The Network Monitoring Role Selection algorithm is used, same thing as for the network monitoring session.
Once the speed test is created in the app, the server will open its port (by default
23999/tcp but configurable in the Agent Settings) and will wait for the client to connect. Once the client is connected, the test will be executed from the client to the server and then from the server to the client.
At any given time, only one speed test can be executed on a monitoring agent. If multiple speed tests are created at the same time, they will be queued until both agents are available.
Agents: Select the agents involved in the speed test. The Network Monitoring Role Selection algorithm is used to select the client and the server.
Maximum Speed: Maximum speeds for both directions in Mbps. If omitted, there is no limit and the speed test will try to transfer as fast as possible.
IP Header DSCP: DSCP value of the IP Header. This can be used to specify a specific QoS Class of Service.
Nb of parallel steams: Number of parallel TCP streams during the test. Default is 4.
Duration (sec): Duration of each speed test transfers (upload & download). The total test duration will be twice the value plus a little overhead.
If the speed test
Duration configured in the speed test settings is higher than the
Organization Speed Test Maximum Duration (configured in the Organization Advanced Settings - (Menu -> Organization Name -> Change Organization's Advanced Parameters), the organization duration will be used. Note that this is also the case for speed tests with public monitoring agents where the public monitoring agent organization can set a maximum duration for the speed tests.
Each agent can have a maximum upload and download limit configured in their Agent Settings. If an agent limit is configured, the speed test limit will be set to the minimum between the speed test setting and the agent setting. This also applies to public monitoring agents that can have limits set.
Each agent has historical maximum download and upload values. These values are a watermark of the maximum results in previous speed tests. At every speed test (manual or scheduled), for each agent, if the speed test result is higher than the stored
Historical Max Speed, the new result will be stored as the new historical max speed. These values are used in the Scheduled Speed Tests and can be erased in the Agent Settings advanced parameters. This is useful if there is a change on the connection speed or the agent is moved to another location.
The maximum speed test is limited by the hardware running the monitoring agent. For example, the hardware appliance X1001 has a fixed limitation to 120 Mbps. For the best performance, we recommend using the Linux Monitoring Agent.
When running a speed test, it's important to ensure that your agents are properly set up to communicate with each other. This means that at least one of the 2 agents is set up to be a server. Read more in the Network Monitoring Role Selection.
The port used for the speed tests is the port configured as Agent Settings under
Incoming TCP/UDP Port. This port must be allowed in the firewall and traffic must be forwarded to the agent if the connection is going through a firewall that does NOT.
No Route to Host error is raised if the speed test process receives an ICMP message that indicates there is no route to host. Usually, this comes from a local firewall that is blocking the connection and sends that ICMP message. Make sure you permit the good TCP port (see above). When this happens, it is possible to capture ICMP messages on the hosts and check the source IP of the ICMP message to identify which network equipment is sending the message.