A lot of you folks have seen the @CloudNativeFdn recent End User Technology Radar that was recently published - radar.cncf.io/2020-09-observ…
It has generated some discussion, but folks I know say takeaways aren't obvious. I've done Observability (TM) for 10 years, here are mine. 1/9
It has generated some discussion, but folks I know say takeaways aren't obvious. I've done Observability (TM) for 10 years, here are mine. 1/9
CNCF published a methodology on how they did this, but they basically asked different companies to classify the observability (and monitoring!) technologies they are using into Adopt, Trial, Assess. radar.cncf.io/how-it-works
Now here's the interesting part: 2/9
Now here's the interesting part: 2/9
Those are current tense verbs and read more like recommendations, when they really shouldn't be. This was a radar, not a prescription (I'm here to give you the prescription ;) It would have been better if they used "Adopted", "Trialed", "Assessed" to be more precise. 3/9
One of the themes is that companies think that running open source such as Prometheus, with a small team is worth the tradeoff to a Saas provider. At a certain scale, this is true. But... 4/9
Another theme snippet is "it may be difficult to switch from one set of tools to another". This is certainly true - even with open source there is technology lock-in. So if you use this radar as a guide, you'll probably find yourself locked into Prometheus or a Saas provider. 5/9
I've worked with several large companies that ended up stuck in Prometheus at 500k metrics per second. Others had proprietary Saas provider statsd extensions which meant they had to renew or retool. So what should you do then? 6/9
Look for Observability vendors that use vendor neutral transports. @circonus in the metrics space ingests JSON telemetry. @honeycombio in the tracing space supports OpenTelemetry. Neither of these were on the radar; probably because respondents were locked in. 7/9
Send your telemetry (logs, metrics, traces) into a pipeline (Kafka is a good one) and transform them on output for vendors. There are a couple observability pipeline solutions out there but IMHO none are production ready at scale. 8/9
Reach out to industry peers on the companies mentioned. We don't have conferences anymore (well, at least those with a hallway track), so you have to reach out, or read their blogs. Remember to ask what pain points they hit; and that their needs are not necessarily yours. 9/9
• • •
Missing some Tweet in this thread? You can try to
force a refresh
