Device Agent Consumption Model
Our Connected Devices product range uses a combination of license model and unit consumption model, owing to its unique use cases. For example, while our Cloud and Enterprise Agents live on selected servers and other static and virtual hardware, our Device Agents have the potential to tap into the very Internet of Things. While most customers currently use Connected Devices to monitor home and office routers, we’ve built our consumption model in a way that can scale as use cases emerge, to keep billing simple even when testing requirements get complex.
Connected Devices Licensing
To perform tests and collect data, every device equipped with a Device Agent needs a license. This license entitles you to the following:
Access to the device within ThousandEyes.
Access to the following network layer tests targeting a private “on-net” test server (see Test Server Methodology for more information about on-net vs. off-net servers):
Download speed
Upload speed
Responsiveness (latency under load)
Latency/loss/disconnections
Jitter
Ping
Cisco Real Speed/instant tests (from within ConstantCare or by API)
The software required to deploy an on-net test server.
Access to view all your device test results on the ThousandEyes platform.
Raw data retention for 30 days.
Aggregated data retention for 90 days.
API configuration.
Access to REST API data.
Access to the software development kits (SDKs) for Cisco Real Speed, mobile apps, and router agents.
Connected Devices Units
To use Connected Devices units, you must first have at least one Connected Devices license. While the Connected Devices license can get your testing regime up and running with our most in-demand network tests, you may wish to expand your test-scape to include some of our more customized tests, or more real-world results from your network layer tests. Tests that require units to run include:
All the network layer tests listed above, but which target an off-net server.
All our other tests, which use different test layers or types. These are summarized as:
Routing layer
DNS layer
Note: All dynamic application test suites consist of multiple tests per suite. For the purposes of unit usage, all the tests that make up a suite are priced as one test execution. See Connected Devices Unit Consumption, below, for more information.
Note: Connected Devices voice layer (simulated voice over IP) tests rely on our UDP jitter framework to run. As such, they incur the cost of the UDP jitter test, which is a network-layer test.
Connected Devices Unit Consumption
As with Cloud and Enterprise Agents, Device Agent units use a fractional model to calculate unit usage. For Cloud and Enterprise Agents, a unit comprises 1000 milli-units, where usage per test is often in the low-digit milli-units. See Calculating Units for more details about the calculation of Cloud and Enterprise Agent units.
For Device Agents, units are consumed at the following rates:
Network-layer off-net tests: a region-specific (US, EU-W, APJC, LATAM, MENA, AUS) fraction of a unit per execution or cycle. Speak to your account manager to find out the specific costs of your off-net tests.
While on-net network-layer tests are free with each license, off-net network-layer tests which use our infrastructure are not. The variable cost of the tests by region reflects the fact that some regions are more expensive for us to maintain than others. This is similar to how tests on Cloud Agents, which we maintain, cost twice as many milli-units as those on Enterprise Agents, which our customers maintain.
Units per execution means one test, executed once, incurs the regional unit cost. These tests are configurable to run at different intervals. Tests that cost per execution are:
Download
Upload
Responsiveness (latency under load)
Units per cycle means you pay a flat regional unit rate for a billing cycle’s worth (one calendar month) of tests. In other words, running these tests over the course of a year would cost the equivalent of 12 individual test executions. Tests that cost per cycle are:
Latency/loss/disconnection: These tests automatically run every 1.5 seconds; they are not configurable to run at any other interval.
Jitter and Ping: These tests are configurable to run at different intervals, but the cost to you is the same per billing cycle regardless of how many or how few tests you run during the cycle.
All other tests: 1/10,000th of a unit per test or test suite. All the below tests are configurable to run at different intervals.
Units per test means one test, executed once, costs 1/10,000 units. Tests that cost 1/10,000 units are:
Traceroute
DNS resolution
Web browsing lite
Generic streaming (HLS/DASH)
Units per test suite means multiple tests, executed once, costs 1/10,000 units. Each test suite is made up of a different number of test endpoints. For example, the gameplay test suite comprises 20+ individual tests to different game endpoints, while the game store test suite comprises just three tests, one to each of the three game store endpoints. You do not need to test to each endpoint within a test suite, but the cost is the same whether you test to one endpoint or 20 endpoints. Test suites that cost 1/10,000 units are:
Gameplay
Video conferencing
Social media
Game store
Video streaming
Content delivery network (CDN)
Example Test Unit Costs
Test | Cost in units per execution |
Download (on-net) | Included |
Upload (on-net) | Included |
Download (off-net) | Variable |
Upload (off-net) | Variable |
Download/upload (off-net) | Variable x 2 |
Ping (off-net testing over 1 billing period) | Variable |
Latency/loss/disconnection (off-net testing over 1 billing period) | Variable |
Video streaming test suite (consisting of up to 3 streaming tests) | 1/10,000 |
Gameplay test suite (consisting of up to 20+ game tests) | 1/10,000 |
DNS resolution | 1/10,000 |
Web browsing lite | 1/10,000 |
Connected Devices Consumption Example
Taxi service WhistleStop want to kit out their fleet of electric cars with Device Agents to ensure each vehicle’s connection to the dispatch center and their customers’ smartphones are always synced in real-time, giving their customers an exceptionally fast, efficient, and data-driven service. The Device Agents can help WhistleStop determine whether the cause of any disruption stems from the internet, the mobile provider (e.g., 4G, 5G), WhistleStop’s own systems, the car itself, or the application being run, greatly enhancing their ability to fix issues promptly. After all, any loss of connection could mean a lost customer. As their fleet are constantly on the move, they need tests to both on-net and off-net servers for network layer tests.
Moreover, they want to provide their customers not only with an exceptional people-delivery service, but one that entertains them along the way. Since they offer their customers free Netflix entertainment on the go via their in-car video screens, they also need to make sure their Netflix streaming connections are top-notch, so they want to employ the video streaming test suite, too.
WhistleStop have 100 electric cars in their fleet, so they buy 100 Device Agent licenses. With these, each car has access to all the on-net network layer tests for free, which they run from their own server installed at WhistleStop HQ. But that’s not enough. Their fleet serve an area of 100 square miles, so they want to test their connectivity from the mobile cell towers their vehicles might be closest to at any given moment, for the most real-world simulation their fleet and customers might be experiencing.
They identify that there are 10 mobile cell towers within the 100-mile radius that their fleet could be connecting to at any given moment. As internet connectivity is critical to their business, they can take advantage of multiple provider coverage to ensure their cars are always connected. They decide that latency/loss and responsiveness are the two network-layer tests they care most about, so they sign up for each Device Agent to conduct the responsiveness test to off-net servers once every five minutes, plus the continuous latency/loss test, both of which incur unit costs.
For the purposes of this example, let's say the off-net network-layer tests cost 1/100 units in their region. The cost of these tests in units for the entire fleet each month is:
((100 cars * (12 tests per hour * 24 hours * 31 days) * 1 responsiveness test (1/100 units)) + 1 latency/loss test (1/100) =
((100 * 8,928) / 100) + 1/100 = 8,928.01 units
Now we can add in the video streaming test suite executions. WhistleStop want to run 1 video streaming test suite execution per car every 5 minutes.
The cost of these video streaming tests in units for the entire fleet each month is:
100 cars * (12 tests per hour * 24 hours * 31 days) * 1 video streaming test suite execution (1/10,000 units)
(100 * 8,928) / 10,000 = 89.28 units
When rounded and added together, WhistleStop would pay for 9,017 units per month for all of their tests for their entire fleet.
Last updated