23:30 A single rogue timestamp which is one hour ahead of the second max timestamp would drop all earlier buckets except one bucket corresponding to this single anomalous value. This is fragile.
Want to know about the best Practices for Real-Time Analytics Architecture on Big Data? Read here: www.graymatter.co.in/real-time-analytics-bigdata-architecture/ Know more: www.graymatter.co.in/real-time-analytics/ Watch here: ua-cam.com/video/lXdYk3hak54/v-deo.html
This is a big disappointment. You cannot stream pipelines built with dataframes. Unified processing framework?? come on! You have to build new versions of all your algorithms so now they can work with a DStream? What a waste of time.
28:45 to 29:40 is the best!!! :D just don't miss that. sets the context
What would be an open source equivalent of DB delta?
Superb Explanation
Very simplified approach of explaining streaming.
Good explanation about streaming..Thanks
Good presentation. Would like to understand more how it could integrate and scale with Apache Kafka.
Hi, please let me know if u have good resources for this. thank you
23:30 A single rogue timestamp which is one hour ahead of the second max timestamp would drop all earlier buckets except one bucket corresponding to this single anomalous value. This is fragile.
Unless there is crazy issue in code manipulating timestamp. It is not a probable scenario to get timestamp ahead of actual unix time.
When the data has entered the dataframe, if the data has been updated or deleted, how can I update or delete it in the dataframe?
A dataframe is immutable. You cannot update a dataframe. You have to create an new dataframe.
Appreciated. Thanks you for a great knowledge share.
Want to know about the best Practices for Real-Time Analytics Architecture on Big Data?
Read here: www.graymatter.co.in/real-time-analytics-bigdata-architecture/
Know more: www.graymatter.co.in/real-time-analytics/
Watch here: ua-cam.com/video/lXdYk3hak54/v-deo.html
How about integrating this with Tensorflow Serving for end to end Analytics paradigm
Google Hydrogen
@@michaelbrenndoerfer9908 lol
This is a big disappointment. You cannot stream pipelines built with dataframes. Unified processing framework?? come on!
You have to build new versions of all your algorithms so now they can work with a DStream? What a waste of time.