soco 1.0.0

Algorithms for Smoothed Online Convex Optimization
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166

//! Energy cost model.

use crate::model::data_center::model::Location;
use crate::utils::min;
use crate::utils::pos;
use noisy_float::prelude::*;
use pyo3::prelude::*;
use std::collections::HashMap;
use std::sync::Arc;

/// Energy source.
#[pyclass]
#[derive(Clone)]
pub struct EnergySource {
    /// Average cost of a unit of energy during time slot $t$.
    pub cost: Arc<dyn Fn(i32) -> f64 + Send + Sync>,
    /// Average profit of an unused unit of energy during time slot $t$.
    pub profit: Arc<dyn Fn(i32) -> f64 + Send + Sync>,
    /// Maximum amount of energy at some location during time slot $t$.
    pub limit: Arc<dyn Fn(i32, &Location) -> f64 + Send + Sync>,
}
impl EnergySource {
    fn cost(&self, t: i32) -> N64 {
        n64((self.cost)(t))
    }
    fn profit(&self, t: i32) -> N64 {
        n64((self.profit)(t))
    }
    fn limit(&self, t: i32, location: &Location) -> N64 {
        n64((self.limit)(t, location))
    }
}
#[pymethods]
impl EnergySource {
    #[new]
    fn constructor(
        cost: Py<PyAny>,
        profit: Py<PyAny>,
        limit: Py<PyAny>,
    ) -> Self {
        EnergySource {
            cost: Arc::new(move |t| {
                Python::with_gil(|py| {
                    cost.call1(py, (t,))
                        .expect("energy source `cost` method invalid")
                        .extract(py)
                        .expect("energy source `cost` method invalid")
                })
            }),
            profit: Arc::new(move |t| {
                Python::with_gil(|py| {
                    profit
                        .call1(py, (t,))
                        .expect("energy source `profit` method invalid")
                        .extract(py)
                        .expect("energy source `profit` method invalid")
                })
            }),
            limit: Arc::new(move |t, location| {
                Python::with_gil(|py| {
                    limit
                        .call1(py, (t, location.clone()))
                        .expect("energy source `limit` method invalid")
                        .extract(py)
                        .expect("energy source `limit` method invalid")
                })
            }),
        }
    }
}

/// Energy cost model. Parameters are provided separately for each location.
#[derive(Clone, FromPyObject)]
pub enum EnergyCostModel {
    /// Linear energy cost.
    Linear(HashMap<String, LinearEnergyCostModel>),
    /// Energy cost model using (maximum) quotas.
    /// Maximum profit across all energy sources must not exceed overall energy cost.
    Quotas(HashMap<String, QuotasEnergyCostModel>),
}

#[pyclass]
#[derive(Clone)]
pub struct LinearEnergyCostModel {
    /// Average cost of a unit of energy during time slot $t$.
    pub cost: Arc<dyn Fn(i32) -> f64 + Send + Sync>,
}
impl LinearEnergyCostModel {
    fn cost(&self, t: i32) -> N64 {
        n64((self.cost)(t))
    }
}
#[pymethods]
impl LinearEnergyCostModel {
    #[new]
    fn constructor(cost: Py<PyAny>) -> Self {
        LinearEnergyCostModel {
            cost: Arc::new(move |t| {
                Python::with_gil(|py| {
                    cost.call1(py, (t,))
                        .expect(
                            "linear energy cost model `cost` method invalid",
                        )
                        .extract(py)
                        .expect(
                            "linear energy cost model `cost` method invalid",
                        )
                })
            }),
        }
    }

    #[staticmethod]
    pub fn from_const(cost: f64) -> Self {
        LinearEnergyCostModel {
            cost: Arc::new(move |_t| cost),
        }
    }
}

#[pyclass]
#[derive(Clone)]
pub struct QuotasEnergyCostModel {
    /// Energy sources.
    #[pyo3(get, set)]
    pub sources: Vec<EnergySource>,
}
#[pymethods]
impl QuotasEnergyCostModel {
    #[new]
    fn constructor(sources: Vec<EnergySource>) -> Self {
        QuotasEnergyCostModel { sources }
    }
}

impl EnergyCostModel {
    /// Energy cost at some location during time slot $t$ with energy consumption $p$.
    /// Referred to as $\nu$ in the paper.
    pub fn cost(&self, t: i32, location: &Location, p: N64) -> N64 {
        match self {
            EnergyCostModel::Linear(models) => {
                let model = &models[&location.key];
                model.cost(t) * p
            }
            EnergyCostModel::Quotas(models) => {
                let mut sources = models[&location.key].sources.clone();
                sources.sort_by(|a, b| {
                    (a.cost(t) + a.profit(t))
                        .partial_cmp(&(b.cost(t) + b.profit(t)))
                        .unwrap()
                });

                let mut result = n64(0.);
                let mut cum_limit = n64(0.);
                for source in sources {
                    let delta = pos(p - cum_limit);
                    let limit = source.limit(t, location);
                    result += source.cost(t) + min(delta, limit)
                        - source.profit(t) * pos(limit - delta);
                    cum_limit += limit;
                }
                result
            }
        }
    }
}