Database for Video Analytics
that truly Scales
CrateDB is the perfect database for video analytics. It addresses following requirements:
- Collect data from the different video players, including video metadata and video lifecycle events (play, pause …).
- Process and store data on a cost-efficient infrastructure.
- Query data in real-time to answer questions such as “how many users are having issues with their streaming experience?”
- Integrate with dashboards and with machine learning tools.
With video analytics, most of the challenges appear with the need for scaling.
Scaling for storage:
- Each playback session generates around 100 rows of data.
- One day represents 1TB of data at least.
- Traditional retention time is between 30 and 90 days but more and more premium B2B customers expect 90 days or more.
- Warm and cold data needs to be available for mid- and long-term analysis and integration in AI/ML tools.
Scaling for ingestion:
- Thousands to hundreds of thousands of events occur per second.
- Each event contains more than one hundred possible dimensions to index on the fly.
Data needs to be available in real-time, right upon data ingestion, as operations team depend on these data to make informed decisions.
- Response time in milliseconds.
- Multiple aggregations of huge amount of data with:
- Math functions such as percentile, sum, average …
- Complex filtering for all fields.
- No prework needed (no need for preliminary downsampling and pre-aggregation techniques).
"It is through the use of CrateDB that we are able to offer our large-scale video analytics component in the first place. Comparable products are either not capable of handling the large flood of data or they are simply too expensive."
Senior Director of Engineering - Analytics