List of Features
Caching Capability
- Metadata Cache (Hive metadata store only)
- Helps during query planning stage (mainly LIST / GET requests)
- Data Cache
- Helps during data read stage
Cache Eviction Methods
When the cache is at capacity, more space is created by evicting data according to the Least Recently Used (LRU) algorithm.
Monitoring Dashboard
Metrics and Grafana template are provided for
- Cluster summary
- Cost saving
- Resource status
Advanced Features
Soft affinity scheduling
This feature helps to ensure the subsequent request is routed to the previous worker to improve on cache hit ratio.
Cache Filter
Sometimes we may not want every single data to go into the cache, and customized admission is necessary. For example
- There is limited space for the cache storage, and there is known “hotness” of the data (or tables).
- Certain data are critical and needs to be fresh always, so we don’t want to cache them. ie. Data lake manifest file.
Cache Filter is such an advanced feature that provides a way for the Admin to define customized logic to control what data can be admitted into the cache.