What is a Rill Compute Unit?
A Rill Compute Unit (RCU) is a usage metering unit that tracks, by the minute, the amount of resources consumed by your Rill cloud service, including compute, memory, disk storage, batch or streaming data ingested.
1. For data ingestion & processing, RCU scale up elastically as you load and transform data into our service.
2. For querying, Rill offers enterprise customers a set of dedicated compute units sized to handle their concurrency requirements across all data sources. Provisioned query RCUs are upscaled or downscaled on a daily basis based on the usage of past week to maintain query performance targets.
Tasks that consume more resources will see more RCU usage compared to tasks that consume fewer resources. While there is no one-to-one mapping between the various resources your service consumes and an RCU, 1 RCU is comparable to the resources used by a task that runs for one hour on 1 vCPU with 4 GB RAM.
How is data size calculated in Rill? When you load data into Rill's service, it is stored in compressed columnar format. Rill charges based on the amount of data stored in Rill's service, after compression. For egress and ingress, actual data transferred over the network is measured. What is the difference between Hot and Cold performance tiers?
Rill provides two performance tiers:
Hot tier is ideal for interactive use-cases which require fast data access. Typically 1-10 sec query response times. Cold tier is ideal for batch reporting use-case where query performance does not matter What performance can I expect from my Hot and Cold tiers? The Hot performance tier provides a target data scan rate up to 1200MBps for every 25GB of data loaded The Cold performance tier provides a target scan rate of up to 100MBps for every 250GB of data loaded How can I increase the performance of a given tier?
Rill will automatically scale RCUs for a given tier to maintain the target scan rates listed above. If higher data scan rates are required, additional RCUs will be provisioned. However, RCU caps placed in a given tier can limit this scaling.
If you are seeing high 95th percentile (P95) query latencies in a given tier, you can take a number of actions:
Increase the RCU cap for a given performance tier along the increments in the table above using the Rill Control Center Reduce the complexity of queries, to limit amount of data being scanned by reducing: the period of data queried, the number of columns accessed in a query, the number of aggregations. Move data from a Cold to a Hot performance tier What are the compute requirements for each performance tier?
For querying, Rill provisions a fixed number of RCUs for each performance tier based on the following factors:
Data Size. Rill will charge a minimum number of compute units per unit of hot storage and cold storage to maintain optimal performance Queries. As the concurrency of data access or complexity of the queries increase Rill will automatically add more RCUs to maintain optimal performance. Performance tier. Hot tier consumes more RCU/TB of stored data than cold tier. This provides a good cost vs performance tradeoff. Cost can be reduced by moving data to the cold tier.
Below are the minimum RCU charged for each performance tier:
8 RCU per 25 GB of data stored
8 RCU per 250 GB of data stored
No RCU consumed. Data is not queryable and can be loaded to either Hot or Cold tier anytime.
The number of RCUs for each performance tier will be added in increments depending on the overall provisioned RCUs.
Overall RCU count Available increment
Up to 64 RCU
Up to 128 RCU
Up to 256 RCU
Over 256 RCU
Provisioned RCUs are upscaled or downscaled on a daily basis based on the usage of past week.
What are the compute requirements for data processing?
For data ingestion & processing, Rill elastically scales up compute slots when you load data into the service.
The number of RCUs consumed depends on the complexity of the ingestion pipeline. A more complex ingestion involving joins will consume more RCUs as compared to a less complex pipeline. How can I estimate my RCU usage?
The best way to get an accurate RCU estimate is to load some sample data into Rill service and track the RCU usage. For estimates on larger projects
contact us for a pricing calculator that reflects the latest volume incentives and discounts. How can I track my RCU usage?
RCU usage and utilization can be viewed in the Rill Control Center. Additionally, Administrators can configure to get Daily, Weekly, Monthly summary reports for RCU usage for visibility.
How can I reduce my RCU consumption?
RCU consumption can be reduced by:
Reducing the ETL complexity Reducing the overall data size stored by: data sampling, increasing data summarization, reprocessing data and dropping less frequently used columns after X days, or reducing the overall data retention period. Moving data to a lower performance tier.