Building Real-Time Analytics Applications Using Apache Pinot

[@migueljimenezZ]

Thank you for this excellent talk!

[@maclovesgeet]

We are using CASSANDRA for metric time series data store. Looking for dashboarding on the top of it. Looked at Superset. BUT superset likes to speak SQL. That research led me to Apache Pinot. How do you compare cassandra vs Pinot for time series data. Numbers. - 1000 metrics, 500k metrics/Minute, 200K dimensions.

[@kishoreg1980]

If you have only metrics and values, Cassandra is good enough but if you have multiple dimensions for each metric, then something like Pinot is a better option.

[@hemanthaugust7217]

​@@kishoreg1980 Do you see any downsides in using a TimeseriesDB such as Prometheus (ignore its alerting & other capabilities, if you don't need them)for this usecase; I agree it's not distributed system. If you have a lot of data, you could explore Grafana Mimir & Grafana UI for dashboarding. Let me know if you see any problems with this solution. It's just 1K metrics and 500k metrics/min datapoints is not a lot. Yes, there are many dimensions to it, and Mimir can shard these and solve it at scale.

[@hemanthaugust7217]

@The Leaf Please explore Grafana Mimir too.

[@ashypeshy]

SUPERB

[@nipuntalukdar]

Great talk.

[@mudunurisrujitha2084]

Is pinot is having any graphql integration point as such?

[@kishoreg1980]

No.

[@mudunurisrujitha2084]

and what is the idea behind choosing the samza as stream processing?

[@kishoreg1980]

Samza was built at LinkedIn. One can use any system for stream processing - Flink, Spark Streaming, etc