The goal of this repository is to have a shared definition of telemetry. By working on a common vocabulary, we can then differentiate tools in an unbiased way. I hope you'll help.
My working definition: Telemetry captures data from multiple sources, aggregates it, and publishes it in a consumable format. What happens after you collect and publish the information is beyond the scope of telemetry.
This gets tricky. I can only start with what I'm curious about, which includes:
Telemetry frameworks are often conflated with entire monitoring solutions. I would like to clarity the differences. To do so, I'm going to focus on only the telemetry portion of lots of tools.
| Tools | Release Status | Collection Model | Written In | API |
|---|---|---|---|---|
| Snap | Beta (0.17.0) | Pull | Go | Yes |
| collectd | Released (5.5) | Pull | C | No |
| Bosun's scollector | Beta (0.5.0) | Pull | Go | Yes |
| OpenTSDB's tcollector | Released (1.3) | Pull | Python | Yes |
| diamond | Released (4.0) | Pull | Python | Yes? |
| Telegraf | Released (1.0 | Pull | Go | No? |
| statsd | Beta (0.8.0) | Push | JavaScript | Yes |
| Ceilometer | Released (7.0.0) | Pull | Python | Yes |
| cAdvisor | Beta (0.24) | Pull | Go | Yes |
| Elastic's metricbeats | Alpha (6.0.0) | Pull | Go | No? |
| metricsd | None | Pull | Go | No |
I will focus only on projects that have public releases from this point forward.
The core functionality of telemetry frameworks.
| Tools | Default Interval | Custom Interval? | Multiple Intervals? | Collect from multiple sources? | Able to Process data? | Publish to multiple sources? |
|---|---|---|---|---|---|---|
| Snap | 1 second | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| collectd | 10 seconds | :white_check_mark: | :x: | :white_check_mark: | :x: | |
| Bosun's scollector | 1 second | :white_check_mark: | ||||
| OpenTSDB's tcollector | 15 seconds | |||||
| Diamond | 300 seconds | |||||
| Telegraf | 10 seconds | |||||
| statsd | N/A | |||||
| Ceilometer | 10 minutes | |||||
| cAdvisor | 1 second | |||||
| Elastic's metricbeats | 10 seconds |
The cost of using these tools, from an operations perspective, is important to consider.
| Tools | New metrics | New collectors | New publishers | Version upgrade | Config change |
|---|---|---|---|---|---|
| Snap | No | No | No | Yes | Yes |
| collectd | Yes | Yes | Yes | Yes | Yes |
| Bosun's scollector | |||||
| OpenTSDB's tcollector | |||||
| Diamond | |||||
| Telegraf | |||||
| statsd | |||||
| Ceilometer | |||||
| cAdvisor | |||||
| Elastic's metricbeats |
The above is a list I'm playing around with. Do you find it helpful? Give it a star to let me know :star: so I know to keep improving it for you. Want to add information? Open a PR following the format above and be sure to support the details with links either in the PR notes or in Notes.md.
What I made:
That I've reviewed:
That I need to explore: