Configuration for autoscaling behavior. More...

#include <autoscaling_policy.h>

Collaboration diagram for kcenon::thread::autoscaling_policy:

Classes
struct	scale_down_config
	Configuration for scale-down triggers. More...

struct	scale_up_config
	Configuration for scale-up triggers. More...

Public Types
enum class	mode { disabled , manual , automatic }
	Autoscaling mode. More...

Public Member Functions
auto	is_valid () const -> bool
	Validates the policy configuration.

Public Attributes
std::size_t	min_workers = 1
	Minimum number of workers (never scale below this)

std::size_t	max_workers = std::thread::hardware_concurrency()
	Maximum number of workers (never scale above this)

scale_up_config	scale_up
	Scale-up trigger configuration.

scale_down_config	scale_down
	Scale-down trigger configuration.

std::size_t	scale_up_increment = 1
	Number of workers to add per scale-up event.

std::size_t	scale_down_increment = 1
	Number of workers to remove per scale-down event.

double	scale_up_factor = 1.5
	Multiplicative factor for scaling (used if use_multiplicative_scaling is true)

bool	use_multiplicative_scaling = false
	Whether to use multiplicative scaling instead of additive.

std::chrono::seconds	scale_up_cooldown {30}
	Minimum time between scale-up events.

std::chrono::seconds	scale_down_cooldown {60}
	Minimum time between scale-down events.

std::chrono::milliseconds	sample_interval {1000}
	Interval between metric samples.

std::size_t	samples_for_decision = 5
	Number of samples to aggregate before making a decision.

mode	scaling_mode = mode::disabled
	Autoscaling mode.

std::function< void(scaling_direction, scaling_reason, std::size_t, std::size_t)>	scaling_callback
	Callback invoked on scaling events Parameters: direction, reason, from_count, to_count.

Detailed Description

Configuration for autoscaling behavior.

This structure defines the policy for automatic scaling of thread pool workers. It includes thresholds for scaling up and down, cooldown periods, and sampling configuration.

Design Principles

Scale-up is triggered by ANY threshold being exceeded (OR logic)
Scale-down requires ALL thresholds to be met (AND logic)
Cooldown periods prevent scaling oscillation
Multiple samples are aggregated before making decisions

Example Configuration

autoscaling_policy policy{
    .min_workers = 2,
    .max_workers = 16,
    .scale_up = {
        .utilization_threshold = 0.8,
        .queue_depth_threshold = 100.0
    },
    .scaling_mode = autoscaling_policy::mode::automatic
};

See also: autoscaler

Definition at line 51 of file autoscaling_policy.h.

Member Enumeration Documentation

◆ mode

enum class kcenon::thread::autoscaling_policy::mode

strong

Autoscaling mode.

Enumerator
disabled	No automatic scaling.
manual	Only scale on explicit trigger.
automatic	Fully automatic scaling.

Definition at line 56 of file autoscaling_policy.h.

        {
            disabled,   
            manual,     
            automatic   
        };

Member Function Documentation

◆ is_valid()

auto kcenon::thread::autoscaling_policy::is_valid ( ) const -> bool

inlinenodiscard

Validates the policy configuration.

Returns: true if valid, false otherwise.

Definition at line 176 of file autoscaling_policy.h.

        {
            if (min_workers == 0)
            {
                return false;
            }
            if (max_workers < min_workers)
            {
                return false;
            }
            if (scale_up.utilization_threshold <= 0.0 || scale_up.utilization_threshold > 1.0)
            {
                return false;
            }
            if (scale_down.utilization_threshold < 0.0 || scale_down.utilization_threshold >= 1.0)
            {
                return false;
            }
            if (scale_down.utilization_threshold >= scale_up.utilization_threshold)
            {
                return false;
            }
            if (scale_up_increment == 0 || scale_down_increment == 0)
            {
                return false;
            }
            if (samples_for_decision == 0)
            {
                return false;
            }
            return true;
        }

References max_workers, min_workers, samples_for_decision, scale_down, scale_down_increment, scale_up, scale_up_increment, kcenon::thread::autoscaling_policy::scale_down_config::utilization_threshold, and kcenon::thread::autoscaling_policy::scale_up_config::utilization_threshold.

Member Data Documentation

◆ max_workers

std::size_t kcenon::thread::autoscaling_policy::max_workers = std::thread::hardware_concurrency()

Maximum number of workers (never scale above this)

Definition at line 108 of file autoscaling_policy.h.

Referenced by kcenon::thread::autoscaler::can_scale_up(), and is_valid().

◆ min_workers

std::size_t kcenon::thread::autoscaling_policy::min_workers = 1

Minimum number of workers (never scale below this)

Definition at line 105 of file autoscaling_policy.h.

Referenced by kcenon::thread::autoscaler::can_scale_down(), and is_valid().

◆ sample_interval

std::chrono::milliseconds kcenon::thread::autoscaling_policy::sample_interval {1000}

Interval between metric samples.

Definition at line 151 of file autoscaling_policy.h.

151{1000};

◆ samples_for_decision

std::size_t kcenon::thread::autoscaling_policy::samples_for_decision = 5

Number of samples to aggregate before making a decision.

Definition at line 154 of file autoscaling_policy.h.

Referenced by is_valid().

◆ scale_down

scale_down_config kcenon::thread::autoscaling_policy::scale_down

Scale-down trigger configuration.

Definition at line 118 of file autoscaling_policy.h.

Referenced by is_valid().

◆ scale_down_cooldown

std::chrono::seconds kcenon::thread::autoscaling_policy::scale_down_cooldown {60}

Minimum time between scale-down events.

Definition at line 144 of file autoscaling_policy.h.

144{60};

Referenced by kcenon::thread::autoscaler::can_scale_down().

◆ scale_down_increment

std::size_t kcenon::thread::autoscaling_policy::scale_down_increment = 1

Number of workers to remove per scale-down event.

Definition at line 128 of file autoscaling_policy.h.

Referenced by is_valid().

◆ scale_up

scale_up_config kcenon::thread::autoscaling_policy::scale_up

Scale-up trigger configuration.

Definition at line 115 of file autoscaling_policy.h.

Referenced by is_valid().

◆ scale_up_cooldown

std::chrono::seconds kcenon::thread::autoscaling_policy::scale_up_cooldown {30}

Minimum time between scale-up events.

Definition at line 141 of file autoscaling_policy.h.

141{30};

Referenced by kcenon::thread::autoscaler::can_scale_up().

◆ scale_up_factor

double kcenon::thread::autoscaling_policy::scale_up_factor = 1.5

Multiplicative factor for scaling (used if use_multiplicative_scaling is true)

Definition at line 131 of file autoscaling_policy.h.

◆ scale_up_increment

std::size_t kcenon::thread::autoscaling_policy::scale_up_increment = 1

Number of workers to add per scale-up event.

Definition at line 125 of file autoscaling_policy.h.

Referenced by is_valid().

◆ scaling_callback

std::function<void(scaling_direction, scaling_reason, std::size_t, std::size_t)> kcenon::thread::autoscaling_policy::scaling_callback

Callback invoked on scaling events Parameters: direction, reason, from_count, to_count.

Definition at line 170 of file autoscaling_policy.h.

◆ scaling_mode

mode kcenon::thread::autoscaling_policy::scaling_mode = mode::disabled

Autoscaling mode.

Definition at line 161 of file autoscaling_policy.h.

◆ use_multiplicative_scaling

bool kcenon::thread::autoscaling_policy::use_multiplicative_scaling = false

Whether to use multiplicative scaling instead of additive.

Definition at line 134 of file autoscaling_policy.h.

The documentation for this struct was generated from the following file:

include/kcenon/thread/scaling/autoscaling_policy.h

Classes

Public Types

Public Member Functions

Public Attributes

Detailed Description

Design Principles

Example Configuration

Member Enumeration Documentation

◆ mode

Member Function Documentation

◆ is_valid()

Member Data Documentation

◆ max_workers

◆ min_workers

◆ sample_interval

◆ samples_for_decision

◆ scale_down

◆ scale_down_cooldown

◆ scale_down_increment

◆ scale_up

◆ scale_up_cooldown

◆ scale_up_factor

◆ scale_up_increment

◆ scaling_callback

◆ scaling_mode

◆ use_multiplicative_scaling