09/10/2022, 4:25 PM
Hello everyone. Recently I tested SigNoz in our company and found it an AMAZING tool to handle APM requirements. I have a question about SigNoz. I noticed that ankitnayan007(maybe @Ankit Nayan? Nice to meet you here! ) said Tempo, a competitor of SigNoz, cannot provide aggregated metrics like RPS and P99 Latency from spans. (link) Tempo released a new feature called metrics generator. The new feature will export histogram and request counts derived from tracing spans into Prometheus and enable deriving RPS and P99 via PromQL. What do you think about this new feature? Will Signoz be superior in some way to Tempo’s metrics generator approach?

Ankit Nayan

09/11/2022, 6:10 AM
Hi @刘瀚骋, When creating metrics from spans, you have to know what sort of metrics you want to extract beforehand. This is not utilising the spans data completely. In SigNoz, you can directly write sql on the raw spans data and plot a chart on runtime which is much more powerful IMO. You don't need to create metrics from spans and query the metrics. Why not just query on the raw spans data.
Eg, filter you traces based on an event inside a span and then get the p99 of those spans which has an attribute
in their attributeMap. For each such query that you want to run on your system, you will first have to create metrics from spans and hence you cannot query on your existing past data as no metrics has been created.