Agents Overview

Agents

How Agents Collect Data

All agents send synthetic probing traffic to internal targets within the network and internet-accessible external targets through the network. Static and cloud agents can also be configured as “managed targets” that respond to probing by other agents. This flexible capability provides the flexibility to deploy agents widely across the network to provide continuous and on-demand insights about network traffic quality to business-critical services inside and outside the network.

Probing:

• SEI users add Agents to probing distributions that define the probing targets, protocols, and intervals.
• Agents receive these instructions automatically from the SEI Controller and begin sending probing traffic and receiving replies from targets at intervals defined in the probing distribution.
• Agents also capture device data like CPU/memory usage, interface loss/errors, and WiFi signal strength (when connected).
• This resulting probing and device data is stored in the agent’s local memory for up to one hour.

SEI Controller operations:

• Every 60 seconds, agents upload stored data to the SEI Controller and check for new or updated probing instructions.
• Locally-stored data is discarded after it is uploaded to the SEI Controller.
• If the agent is unable to reach the SEI Controller for more than one hour, probing continues, and the buffered data older than one hour are overwritten with each new minute of data. Only the last hour of test data is stored locally.
• When the agent reconnects to the SEI Controller, the last hour of test data is uploaded to the time series database and erased from local memory.

On-demand Troubleshooting:

• Because agents are continuously connected to the SEI Controller, troubleshooting commands and the resulting test data are streamed between that agent and SEI Controller in real-time.

Metrics By Agent

All agents provide time series metrics and on-demand troubleshooting data.

Following is a quick reference to the metrics provide by agent type. See the Metrics article for details about each type of metric.

• WiFi Signal Strength:
  • WiFi metrics are available for agents connected to the controller through a WiFi access point.
  • While cloud Agents have the same capabilities as Static Agents, it’s unlikely that cloud-hosting environments would provide a way to connect the agent to a WiFi access point.
• Packet Capture (PCAP) is not supported by mobile agents.

Metric Type	Metric	Static and Cloud Agents	Mobile Agents
Probing metrics (time series)	Latency
	Jitter
	Loss
	HTTP Availability
	HTTP Request Response Time
	Path Discovery (ASN, hops latency)
	Speed Test
Device metrics (time series)	CPU and Memory Usage
	Interface Loss TX and RX
	Interface Errors TX and RX
	WiFi Signal Strength (when connected)
On-demand troubleshooting (streaming)	Packet Capture		–
	Ping Test
	DNS Lookup
	HTTP Test
	Path Discovery

Path Discovery

Agents perform path discovery when the agent is added to a probing distribution with path discovery enabled.

Enable path discovery and set the path probing interval in Add/Edit Probing Distribution modal.

Once enabled, the path probing interval and “enabled” toggle are shown on the probing distribution page and can be toggled on and off.

Agents perform a path probing (also referred to as traceroute) to create path discovery visualizations of the hops between the agent and the target. The path discovery visualization shows the IP and latency of each hop and the carrier ASN in use at the time.

For agents in a probing distribution with path discovery enabled, clicking a target plotline opens the path visualization for the paths probed during the time range selected.

Learn more about path discovery metrics and path probing methods in the Metrics article

**Path Discovery Triggers**

Path discovery is the product of path probing that is triggered in four ways:

1. Path Probing Interval

• A path probing interval of one, six, 12, or 24 hours is set when path discovery is enabled in a probing distribution.
• The agent performs path discovery for this interval until path discovery is disabled in the probing distribution, the agent is removed from the probing distribution, or the agent is offline.

2. Latency Change

• Path Probing is triggered automatically when the latency between the agent and target increases.
• This feature enables the agent to capture hop metrics during meaningful events between the periodic path discovery intervals.
• Each time an agent probes a target, a latency value is created. When a new latency value is created, the agent compares it to the lowest latency value from the last two minutes. Path probing is triggered when the new latency value increases per the deviation thresholds below:

Normal Latency	Deviation Threshold
100 ms	+/- 10 ms
35 ms to 99.99 ms	+/- 5 ms
< 35 ms	+/- 3 ms

The following example shows how comparison windows are used to automatically trigger path probing when latency increases.

• The probing distribution for the example agent/target combination has a 30-second probing interval resulting in two latency values per minute.
• Minute 3:
  • The first new latency value is 41 ms.
  • The minimum latency value for the previous two-minute comparison window is 25 ms.
  • The agent compares these values and triggers path probing because the comparison minimum latency is <35 ms, and the new latency value increases by 3 ms or more.
  • After 30 seconds, the following new latency value is 42 ms. While the comparison window has advanced by 30 seconds, the comparison value remains 25 ms, and path probing is triggered again.
• Minute 4:
  • The first new latency value is 43 ms; the comparison window minimum value remains unchanged. Path probing is triggered again.
  • After 30 seconds, the second latency value or the minute is 42ms.
  • The two-minute comparison window has advanced, the minimum latency is now 40 ms, and path probing is not triggered.

• Path discovery is the product of path probing
  • All SEI metrics are created from all probing data collected during the probing interval.
  • Path discovery metrics and resulting visualizations are also the product of path-probing data collected during the probing interval. Therefore, path probing may occur twice a minute, but only one set of path discovery metrics is created.
• Two-minute comparison windows are typical but not absolute.
  • Most probing distributions use a probing interval of one minute or less, resulting in at least one latency value per minute.
  • If the probing interval is greater than one minute, the last two values are used for comparison.

3. Target IP change detected

• Path probing is triggered when a change to the target’s IP is detected.
• For example, HTTP Targets like SaaS services typically use IP pools to optimize service availability. The agent recognizes these changes and conducts path probing to each IP.

4. On-demand Path Discovery:

• On-demand path discovery conducts path probing to existing targets or any reachable IP/URL.
• When triggered, path probing data streams from the agent in real-time.
• Results are not stored in the time series database and do not appear in the path discovery visualization.

Path probing protocols

The protocol used for path probing vary by path probing trigger, target protocol, and agent type.

	Target Protocol	Static and Cloud Agents (Linux-based)	Mobile Agent for macOS	Mobile Agent for Windows
Path Discovery Interval	HTTP	TCP	UDP	ICMP
	UDP	UDP	UDP	ICMP
	ICMP	ICMP	ICMP	ICMP
Latency Change & Target IP Change Detected	HTTP	TCP	UDP	ICMP
	UDP	UDP	UDP	ICMP
	ICMP	ICMP & UDP	ICMP	ICMP
On-Demand	Any	ICMP, UDP, TCP	ICMP, UDP, TCP	ICMP

Path probing attempts

The number of attempts differs by path probing trigger.

	Number of Attempts
Path Discovery Interval	10
Latency Change & Target IP Change Detected	3
On-demand	1-10 set by the user, 3 is the default setting