[poolifier.git] / docs / worker-choice-strategies.md

# Worker choice strategies

All duration or timestamp are expressed in milliseconds.

## Table of contents

- [Strategies](#strategies)
  - [Fair share](#fair-share)
  - [Weighted round robin](#weighted-round-robin)
  - [Interleaved weighted round robin](#interleaved-weighted-round-robin)
- [Statistics](#statistics)
  - [Simple moving median](#simple-moving-median)

## Strategies

### Fair share

Its goal is to distribute the load evenly across all workers. To achieve this, the strategy keeps track of the simple moving average task execution time for each worker and assigns the next task to the worker with the lowest task end prediction time: `task_end_prediction = max(current_time, task_end_prediction) + simple_moving_average_task_execution_time`.  
By default, the strategy uses the simple moving average task execution time for each worker but it can be configured to use the simple moving average task event loop utilization (ELU) active time instead.

### Weighted round robin

The worker weights are maximum tasks execution time, once the worker has reached its maximum tasks execution time, the next task is assigned to the next worker. The worker default weights are the same for all workers and is computed given the CPU cores speed and theirs numbers.

### Interleaved weighted round robin

The worker weights are maximum tasks execution time. The rounds are the deduplicated worker weights.  
During a round, if worker weight is inferior to the current round, the next task is assigned to the next worker. Once all workers have been assigned a task, the next round starts.  
The worker default weights are the same for all workers and is computed given the CPU cores speed and theirs numbers. So the default rounds consists of a unique worker weight.

## Statistics

Worker choice strategies enable only the statistics that are needed to choose the next worker to avoid unnecessary overhead.

### Simple moving median

Strategies using the average task execution time for each worker can use the simple moving median instead. Simple moving median is more robust to outliers and can be used to avoid assigning tasks to workers that are currently overloaded. Simple moving median usage introduces a small overhead: measurement history must be kept for each worker and the simple moving median must be recomputed each time a task has finished.
Commit	Line	Data
a47027a0 JB	1	# Worker choice strategies
	2
	3	All duration or timestamp are expressed in milliseconds.
	4
	5	## Table of contents
	6
	7	- [Strategies](#strategies)
	8	- [Fair share](#fair-share)
	9	- [Weighted round robin](#weighted-round-robin)
	10	- [Interleaved weighted round robin](#interleaved-weighted-round-robin)
	11	- [Statistics](#statistics)
dc021bcc	12	- [Simple moving median](#simple-moving-median)
a47027a0 JB	13
	14	## Strategies
	15
	16	### Fair share
	17
dc021bcc JB	18	Its goal is to distribute the load evenly across all workers. To achieve this, the strategy keeps track of the simple moving average task execution time for each worker and assigns the next task to the worker with the lowest task end prediction time: `task_end_prediction = max(current_time, task_end_prediction) + simple_moving_average_task_execution_time`.
dc021bcc JB	19	By default, the strategy uses the simple moving average task execution time for each worker but it can be configured to use the simple moving average task event loop utilization (ELU) active time instead.
a47027a0 JB	20
	21	### Weighted round robin
	22
	23	The worker weights are maximum tasks execution time, once the worker has reached its maximum tasks execution time, the next task is assigned to the next worker. The worker default weights are the same for all workers and is computed given the CPU cores speed and theirs numbers.
	24
	25	### Interleaved weighted round robin
	26
	27	The worker weights are maximum tasks execution time. The rounds are the deduplicated worker weights.
	28	During a round, if worker weight is inferior to the current round, the next task is assigned to the next worker. Once all workers have been assigned a task, the next round starts.
	29	The worker default weights are the same for all workers and is computed given the CPU cores speed and theirs numbers. So the default rounds consists of a unique worker weight.
	30
	31	## Statistics
	32
	33	Worker choice strategies enable only the statistics that are needed to choose the next worker to avoid unnecessary overhead.
	34
dc021bcc	35	### Simple moving median
a47027a0	36
dc021bcc	37	Strategies using the average task execution time for each worker can use the simple moving median instead. Simple moving median is more robust to outliers and can be used to avoid assigning tasks to workers that are currently overloaded. Simple moving median usage introduces a small overhead: measurement history must be kept for each worker and the simple moving median must be recomputed each time a task has finished.