NCS/age_cache_simulation
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

feat(eta_calculation): Implementing "Joint Caching and Freshness Optimization" (in progress)

Browse Source 
Signed-off-by: Tuan-Dat Tran <tuan-dat.tran@tudattr.dev>
This commit is contained in:
Tuan-Dat Tran
2024-12-02 18:07:08 +01:00
parent b2cc80bb09
commit 4ea5505130
7 changed files with 47164 additions and 42 deletions
Download Patch File Download Diff File Expand all files Collapse all files

473
01_nb_cncf_optimization/.ipynb_checkpoints/gen_nb_cost_optimization-checkpoint.ipynb Normal file
View File

@@ -0,0 +1,473 @@
{
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"id": "ab5cd7d1-1a57-46fc-8282-dae0a6cc2944",
"metadata": {},
"outputs": [],
"source": [
"import matplotlib.pyplot as plt\n",
"import numpy as np\n",
"import random"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "3d1ad0b9-f6a8-4e98-84aa-6e02e4279954",
"metadata": {},
"outputs": [],
"source": [
"DATABASE_OBJECT_COUNT = 100\n",
"CACHE_SIZE = DATABASE_OBJECT_COUNT/2\n",
"ZIPF_CONSTANT = 2\n",
"\n",
"CACHE_MISS_COST = 2\n",
"CACHE_REFRESH_COST = 1\n",
"\n",
"SEED = 42\n",
"np.random.seed(SEED)\n",
"random.seed(SEED)\n",
"\n",
"LAMBDA_VALUES = np.array([np.random.zipf(ZIPF_CONSTANT) for i in np.arange(1, DATABASE_OBJECT_COUNT + 1,1)])"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "9cc83cf6-5c78-4f0d-b7cb-08cdb80c362e",
"metadata": {},
"outputs": [],
"source": [
"# LAMBDA_VALUES = np.array([0.03, 0.04,0.05,0.06,0.07,1,1.1,1.2,1.3,1.4,1.5])\n",
"# DATABASE_OBJECT_COUNT = len(LAMBDA_VALUES)\n",
"# CACHE_SIZE = 4.4\n",
"# CACHE_MISS_COST = 7\n",
"# CACHE_REFRESH_COST = 1"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "3dc07233-0b56-4fee-a93b-212836c18b42",
"metadata": {},
"outputs": [],
"source": [
"db_object_count = DATABASE_OBJECT_COUNT\n",
"cache_sz = CACHE_SIZE\n",
"\n",
"lambda_vals = LAMBDA_VALUES\n",
"c_f = CACHE_MISS_COST\n",
"c_delta = CACHE_REFRESH_COST"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "5a27d416-8f98-4814-af9e-6c6bef95f4ef",
"metadata": {},
"outputs": [],
"source": [
"def eta_star(db_object_count, c_f, cache_sz, c_delta, lambda_vals):\n",
" num = (db_object_count * c_f - cache_sz * c_delta)\n",
" denom = np.sum(1.0/lambda_vals)\n",
" return max(0, num/denom)"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "6276a9ce-f839-4fe6-90f2-2195cf065fc8",
"metadata": {},
"outputs": [],
"source": [
"def h_i_star(c_f, eta, lambda_vals, c_delta):\n",
" optimized_hitrate = (c_f - (eta/lambda_vals)) / c_delta\n",
" return optimized_hitrate"
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "dcd31a8c-6864-4b9a-8bb3-998f0c32baf6",
"metadata": {},
"outputs": [],
"source": [
"def get_index_of_furthest_hitrate_from_boundary(hitrates):\n",
" local_hitrates = hitrates[(hitrates < 0) | (hitrates > 1)]\n",
" smallest_delta = np.abs(np.min(local_hitrates))\n",
" biggest_delta = np.max(local_hitrates) - 1\n",
" if smallest_delta > biggest_delta:\n",
" index = np.where(hitrates == np.min(local_hitrates))[0][0]\n",
" return index\n",
" else:\n",
" index = np.where(hitrates == np.max(local_hitrates))[0][0]\n",
" return index"
]
},
{
"cell_type": "code",
"execution_count": 18,
"id": "55b251f8-97ca-49a8-9ec6-be77dc1e49b2",
"metadata": {
"scrolled": true
},
"outputs": [],
"source": [
"\"\"\"\n",
"Perform theoretical optimization to compute optimal hit probabilities.\n",
"\n",
"Parameters:\n",
"- lambda_vals (numpy array): Request rates for each item.\n",
"- B (float): Total cache size.\n",
"- c_f (float): Fetching linear cost (cache miss cost).\n",
"- c_delta (float): Age linear cost.\n",
"\n",
"Returns:\n",
"- h_opt (numpy array): Optimal hit probabilities for each item.\n",
"\"\"\"\n",
"optimized_hitrates = np.zeros(DATABASE_OBJECT_COUNT)\n",
"differenc_set = np.arange(DATABASE_OBJECT_COUNT)\n",
"fix_i = []\n",
"current_db_objects = DATABASE_OBJECT_COUNT\n",
"current_cache_size = CACHE_SIZE\n",
"\n",
"while True:\n",
" if current_db_objects == 0:\n",
" # Handle special case: no items left to optimize\n",
" if current_cache_size > 0:\n",
" # Redistribute unused cache size among items with zero hit probability\n",
" differenc_set = np.where(optimized_hitrates == 0)[0]\n",
" fix_i = np.setdiff1d(np.arange(DATABASE_OBJECT_COUNT), differenc_set)\n",
" current_db_objects = len(differenc_set)\n",
" continue\n",
" else:\n",
" optimized_hitrates[differenc_set] = 0\n",
" break\n",
" # Compute Lagrangian multiplier and optimal hit probabilities\n",
" mu = max(0, (current_db_objects * c_f - current_cache_size * c_delta) / np.sum(1.0 / lambda_vals[differenc_set]))\n",
" eta = eta_star(current_db_objects, c_f, current_cache_size, c_delta, lambda_vals[differenc_set])\n",
" assert(mu == eta)\n",
" optimized_hitrates[differenc_set] = (c_f - mu / lambda_vals[differenc_set]) / c_delta\n",
" # print(optimized_hitrates)\n",
" # Handle the case where mu < 0\n",
" if mu < 0:\n",
" current_cache_size = current_db_objects * c_f / c_delta # Adjust cache size for next iteration\n",
" continue\n",
" # Check for constraint violations\n",
" larger_i = np.where(optimized_hitrates > 1)[0] # h > 1\n",
" smaller_i = np.where(optimized_hitrates < 0)[0] # h < 0\n",
" # If no violations, optimization is complete\n",
" break_con = len(smaller_i) == 0 and len(larger_i) == 0\n",
" break_con1 = len((optimized_hitrates[differenc_set])[((optimized_hitrates[differenc_set]) < 0) | ((optimized_hitrates[differenc_set])> 1)]) == 0\n",
" assert(break_con == break_con1)\n",
" if break_con:\n",
" break\n",
" # Find the furthest violating item\n",
" min_viol, min_viol_i = (0, -1)\n",
" if len(smaller_i) > 0:\n",
" min_viol_i = np.argmin(optimized_hitrates)\n",
" min_viol = optimized_hitrates[min_viol_i]\n",
" max_viol, max_viol_i = (0, -1)\n",
" if len(larger_i) > 0:\n",
" larger = optimized_hitrates - 1\n",
" max_viol_i = np.argmax(larger)\n",
" max_viol = larger[max_viol_i]\n",
" # Compare the furthest violations and adjust accordingly\n",
" viol_i = min_viol_i\n",
" min_viol_flag = True # True if furthest is from the left boundary\n",
" if max_viol > abs(min_viol):\n",
" viol_i = max_viol_i\n",
" min_viol_flag = False \n",
" index = get_index_of_furthest_hitrate_from_boundary(optimized_hitrates)\n",
" if viol_i != index:\n",
" print(optimized_hitrates[viol_i])\n",
" print(optimized_hitrates[index])\n",
" assert(viol_i == index)\n",
" if min_viol_flag:\n",
" optimized_hitrates[viol_i] = 0\n",
" else:\n",
" optimized_hitrates[viol_i] = min(1, current_cache_size)\n",
"\n",
" # Update parameters for next iteration\n",
" current_cache_size =- optimized_hitrates[viol_i]\n",
" fix_i.append(viol_i)\n",
" differenc_set = np.setdiff1d(np.arange(DATABASE_OBJECT_COUNT), fix_i)\n",
" current_db_objects = DATABASE_OBJECT_COUNT - len(fix_i)"
]
},
{
"cell_type": "code",
"execution_count": 9,
"id": "efa16eaf-a10b-4927-99cd-190e2ffe1d1e",
"metadata": {},
"outputs": [],
"source": [
"a = optimized_hitrates\n",
"b = differenc_set"
]
},
{
"cell_type": "code",
"execution_count": 10,
"id": "0e21c26f-058a-4e56-a5ad-1c47bf28656c",
"metadata": {
"scrolled": true
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"All values optimized.\n"
]
}
],
"source": [
"\"\"\"\n",
"Perform theoretical optimization to compute optimal hit probabilities.\n",
"\n",
"Parameters:\n",
"- lambda_vals (numpy array): Request rates for each item.\n",
"- B (float): Total cache size.\n",
"- c_f (float): Fetching linear cost (cache miss cost).\n",
"- c_delta (float): Age linear cost.\n",
"\n",
"Returns:\n",
"- h_opt (numpy array): Optimal hit probabilities for each item.\n",
"\"\"\"\n",
"optimized_hitrates = np.zeros(DATABASE_OBJECT_COUNT)\n",
"differenc_set = np.arange(DATABASE_OBJECT_COUNT)\n",
"fix_i = []\n",
"current_db_objects = DATABASE_OBJECT_COUNT\n",
"current_cache_size = CACHE_SIZE\n",
"\n",
"while True:\n",
" if current_db_objects == 0:\n",
" # Handle special case: no items left to optimize\n",
" if current_cache_size > 0:\n",
" # Redistribute unused cache size among items with zero hit probability\n",
" differenc_set = np.where(optimized_hitrates == 0)[0]\n",
" fix_i = np.setdiff1d(np.arange(DATABASE_OBJECT_COUNT), differenc_set)\n",
" current_db_objects = len(differenc_set)\n",
" continue\n",
" else:\n",
" optimized_hitrates[differenc_set] = 0\n",
" break\n",
" # Compute Lagrangian multiplier and optimal hit probabilities\n",
" eta = eta_star(current_db_objects, c_f, current_cache_size, c_delta, lambda_vals[differenc_set])\n",
" optimized_hitrates[differenc_set] = (c_f - eta / lambda_vals[differenc_set]) / c_delta\n",
"\n",
" if mu < 0:\n",
" current_cache_size = current_db_objects * c_f / c_delta # Adjust cache size for next iteration\n",
" continue\n",
" \n",
" if len((optimized_hitrates[differenc_set])[((optimized_hitrates[differenc_set]) < 0) | ((optimized_hitrates[differenc_set])> 1)]) == 0:\n",
" print(\"All values optimized.\")\n",
" break\n",
" max_outbound_index = get_index_of_furthest_hitrate_from_boundary(optimized_hitrates)\n",
" optimized_hitrates[max_outbound_index] = (1 if optimized_hitrates[max_outbound_index] > 1 else 0)\n",
"\n",
" current_cache_size =- optimized_hitrates[max_outbound_index]\n",
" fix_i.append(max_outbound_index)\n",
" differenc_set = np.setdiff1d(np.arange(DATABASE_OBJECT_COUNT), fix_i)\n",
" current_db_objects = DATABASE_OBJECT_COUNT - len(fix_i)"
]
},
{
"cell_type": "code",
"execution_count": 11,
"id": "4f64253f-b389-4be9-b403-08027d480121",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
" 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
" 0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0.,\n",
" 1., 0., 0., 0., 0., 0., 1., 0., 0., 1., 0., 0., 0., 0., 1., 0., 1.,\n",
" 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0.,\n",
" 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0.])"
]
},
"execution_count": 11,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"optimized_hitrates"
]
},
{
"cell_type": "code",
"execution_count": 12,
"id": "17d818db-ec88-4c26-92af-6d74862525d9",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([ 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0.43902439, 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0.43902439, 0. , 0. ,\n",
" 0. , 0.04878049, 0. , 0.04878049, 0. ,\n",
" 0. , 0. , 0. , -0. , 0. ,\n",
" 0.04878049, 0.04878049, 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0.43902439, 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" -0. , 0. , 0. , 0. , 0. ,\n",
" -0. , 0. , 1. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , -0. , 0. ,\n",
" 0. , 0.43902439, 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0.04878049, 0. ])"
]
},
"execution_count": 12,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"a"
]
},
{
"cell_type": "code",
"execution_count": 13,
"id": "791b3f96-527a-489e-970e-c92ec950177f",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([ 6, 27, 31, 33, 40, 41, 50, 81, 98])"
]
},
"execution_count": 13,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"b"
]
},
{
"cell_type": "code",
"execution_count": 14,
"id": "c22fa973-432a-4c05-89bf-2a6ea82ae3d2",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([ 6, 27, 50, 81])"
]
},
"execution_count": 14,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"differenc_set"
]
},
{
"cell_type": "code",
"execution_count": 15,
"id": "898e1266-5aaa-46f4-ac0f-c7807ac2b6bb",
"metadata": {},
"outputs": [],
"source": [
"db_object_count = DATABASE_OBJECT_COUNT\n",
"cache_sz = CACHE_SIZE\n",
"loop_lambda = lambda_vals\n",
"\n",
"non_optimized_values = np.arange(db_object_count)"
]
},
{
"cell_type": "code",
"execution_count": 16,
"id": "8cc9b8a9-f7ae-48fc-adfb-ac4b7a4998f1",
"metadata": {},
"outputs": [],
"source": [
"db_object_count = DATABASE_OBJECT_COUNT\n",
"cache_sz = CACHE_SIZE\n",
"loop_lambda = lambda_vals\n",
"\n",
"optimized_hitrate = np.zeros(db_object_count)\n",
"non_optimized_values = np.arange(db_object_count)\n",
"optimized_value = {}\n",
"\n",
"eta = eta_star(db_object_count, c_f, cache_sz, c_delta, loop_lambda[non_optimized_values])\n",
"optimized_hitrate[non_optimized_values] = h_i_star(c_f, eta, loop_lambda[non_optimized_values], c_delta)\n",
"\n",
"max_outbound_index = get_index_of_furthest_hitrate_from_boundary(optimized_hitrate)\n",
"optimized_value[max_outbound_index] = (1 if optimized_hitrate[max_outbound_index] > 1 else 0)"
]
},
{
"cell_type": "code",
"execution_count": 17,
"id": "cbcf3592-fcf2-4f54-a3cd-761097c12972",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"{67: 1}"
]
},
"execution_count": 17,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"optimized_value"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "db732331-1d09-45b7-915c-73daa270b5e2",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "graphs",
"language": "python",
"name": "graphs"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.12.7"
}
},
"nbformat": 4,
"nbformat_minor": 5
}

247
01_nb_cncf_optimization/.ipynb_checkpoints/nb_cost_optimization-checkpoint.ipynb
View File

@@ -1,6 +1,251 @@
{
"cells": [],
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"id": "ab5cd7d1-1a57-46fc-8282-dae0a6cc2944",
"metadata": {},
"outputs": [],
"source": [
"import matplotlib.pyplot as plt\n",
"import numpy as np\n",
"import random"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "3d1ad0b9-f6a8-4e98-84aa-6e02e4279954",
"metadata": {},
"outputs": [],
"source": [
"DATABASE_OBJECT_COUNT = 100\n",
"CACHE_SIZE = DATABASE_OBJECT_COUNT/2\n",
"ZIPF_CONSTANT = 2\n",
"\n",
"CACHE_MISS_COST = 1\n",
"CACHE_REFRESH_COST = 1\n",
"\n",
"SEED = 42\n",
"np.random.seed(SEED)\n",
"random.seed(SEED)\n",
"\n",
"LAMBDA_VALUES = np.array([np.random.zipf(ZIPF_CONSTANT) for i in np.arange(1, DATABASE_OBJECT_COUNT + 1,1)])"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "9cc83cf6-5c78-4f0d-b7cb-08cdb80c362e",
"metadata": {},
"outputs": [],
"source": [
"# LAMBDA_VALUES = np.array([0.03, 0.04,0.05,0.06,0.07,1,1.1,1.2,1.3,1.4,1.5])\n",
"# DATABASE_OBJECT_COUNT = len(LAMBDA_VALUES)\n",
"# CACHE_SIZE = 4.4\n",
"# CACHE_MISS_COST = 7\n",
"# CACHE_REFRESH_COST = 1"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "3dc07233-0b56-4fee-a93b-212836c18b42",
"metadata": {},
"outputs": [],
"source": [
"db_object_count = DATABASE_OBJECT_COUNT\n",
"cache_sz = CACHE_SIZE\n",
"\n",
"lambda_vals = LAMBDA_VALUES\n",
"c_f = CACHE_MISS_COST\n",
"c_delta = CACHE_REFRESH_COST"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "5a27d416-8f98-4814-af9e-6c6bef95f4ef",
"metadata": {},
"outputs": [],
"source": [
"def eta_star(db_object_count, c_f, cache_sz, c_delta, lambda_vals):\n",
" num = (db_object_count * c_f - cache_sz * c_delta)\n",
" denom = np.sum(1.0/lambda_vals)\n",
" return max(0, num/denom)"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "6276a9ce-f839-4fe6-90f2-2195cf065fc8",
"metadata": {},
"outputs": [],
"source": [
"def h_i_star(c_f, eta, lambda_vals, c_delta):\n",
" optimized_hitrate = (c_f - (eta/lambda_vals)) / c_delta\n",
" return optimized_hitrate"
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "dcd31a8c-6864-4b9a-8bb3-998f0c32baf6",
"metadata": {},
"outputs": [],
"source": [
"def get_index_of_furthest_hitrate_from_boundary(hitrates):\n",
" local_hitrates = hitrates[(hitrates < 0) | (hitrates > 1)]\n",
" smallest_delta = np.abs(np.min(local_hitrates))\n",
" biggest_delta = np.max(local_hitrates) - 1\n",
" if smallest_delta > biggest_delta:\n",
" index = np.where(hitrates == np.min(local_hitrates))[0][0]\n",
" return index\n",
" else:\n",
" index = np.where(hitrates == np.max(local_hitrates))[0][0]\n",
" return index"
]
},
{
"cell_type": "code",
"execution_count": 14,
"id": "ccd4b95d-1cdd-4c99-a22e-4b31338993cf",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([0.30256805, 0.76752268, 0.30256805, 0.30256805, 0.65128403,\n",
" 0.30256805, 0.86051361, 0.30256805, 0.30256805, 0.30256805,\n",
" 0.65128403, 0.30256805, 0.30256805, 0.30256805, 0.65128403,\n",
" 0.65128403, 0.30256805, 0.30256805, 0.76752268, 0.30256805,\n",
" 0.30256805, 0.30256805, 0.30256805, 0.65128403, 0.30256805,\n",
" 0.30256805, 0.30256805, 0.86051361, 0.30256805, 0.30256805,\n",
" 0.30256805, 0.82564201, 0.30256805, 0.82564201, 0.30256805,\n",
" 0.30256805, 0.30256805, 0.76752268, 0.91282101, 0.30256805,\n",
" 0.82564201, 0.82564201, 0.65128403, 0.30256805, 0.30256805,\n",
" 0.30256805, 0.93025681, 0.30256805, 0.30256805, 0.30256805,\n",
" 0.86051361, 0.92250756, 0.30256805, 0.30256805, 0.30256805,\n",
" 0.30256805, 0.30256805, 0.95897459, 0.65128403, 0.30256805,\n",
" 0.97317569, 0.30256805, 0.30256805, 0.65128403, 0.30256805,\n",
" 0.93025681, 0.30256805, 0.98989229, 0.30256805, 0.30256805,\n",
" 0.65128403, 0.30256805, 0.30256805, 0.30256805, 0.76752268,\n",
" 0.65128403, 0.65128403, 0.76752268, 0.95350454, 0.30256805,\n",
" 0.30256805, 0.86051361, 0.65128403, 0.30256805, 0.30256805,\n",
" 0.65128403, 0.30256805, 0.65128403, 0.30256805, 0.30256805,\n",
" 0.65128403, 0.65128403, 0.76752268, 0.30256805, 0.65128403,\n",
" 0.30256805, 0.30256805, 0.98115049, 0.82564201, 0.65128403])"
]
},
"execution_count": 14,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"eta = eta_star(db_object_count, c_f, cache_sz, c_delta, lambda_vals)\n",
"optimized_hitrates = (c_f - eta / lambda_vals) / c_delta\n",
"optimized_hitrates"
]
},
{
"cell_type": "code",
"execution_count": 8,
"id": "0e21c26f-058a-4e56-a5ad-1c47bf28656c",
"metadata": {
"scrolled": true
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"All values optimized.\n"
]
}
],
"source": [
"\"\"\"\n",
"Perform theoretical optimization to compute optimal hit probabilities.\n",
"\n",
"Parameters:\n",
"- lambda_vals (numpy array): Request rates for each item.\n",
"- B (float): Total cache size.\n",
"- c_f (float): Fetching linear cost (cache miss cost).\n",
"- c_delta (float): Age linear cost.\n",
"\n",
"Returns:\n",
"- h_opt (numpy array): Optimal hit probabilities for each item.\n",
"\"\"\"\n",
"optimized_hitrates = np.zeros(DATABASE_OBJECT_COUNT)\n",
"current_db_object_count = DATABASE_OBJECT_COUNT\n",
"current_cache_size = CACHE_SIZE\n",
"\n",
"differenc_set = np.arange(DATABASE_OBJECT_COUNT)\n",
"fix_i = []\n",
"\n",
"while True:\n",
" if current_db_object_count == 0:\n",
" print(\"No objects left to optimize.\")\n",
" if current_cache_size > 0:\n",
" # Redistribute unused cache size among items with zero hit probability\n",
" differenc_set = np.where(optimized_hitrates == 0)[0]\n",
" fix_i = np.setdiff1d(np.arange(DATABASE_OBJECT_COUNT), differenc_set)\n",
" current_db_object_count = len(differenc_set)\n",
" continue\n",
" else:\n",
" optimized_hitrates[differenc_set] = 0\n",
" break\n",
" # Compute Lagrangian multiplier and optimal hit probabilities\n",
" eta = eta_star(current_db_object_count, c_f, current_cache_size, c_delta, lambda_vals[differenc_set])\n",
" optimized_hitrates[differenc_set] = (c_f - eta / lambda_vals[differenc_set]) / c_delta\n",
"\n",
" if eta < 0:\n",
" print(\"eta was negative.\")\n",
" current_cache_size = current_db_object_count * c_f / c_delta # Adjust cache size for next iteration\n",
" continue\n",
" \n",
" if len((optimized_hitrates[differenc_set])[((optimized_hitrates[differenc_set]) < 0) | ((optimized_hitrates[differenc_set])> 1)]) == 0:\n",
" print(\"All values optimized.\")\n",
" break\n",
" \n",
" max_outbound_index = get_index_of_furthest_hitrate_from_boundary(optimized_hitrates)\n",
" optimized_hitrates[max_outbound_index] = (1 if optimized_hitrates[max_outbound_index] > 1 else 0)\n",
"\n",
" current_cache_size =- optimized_hitrates[max_outbound_index]\n",
" fix_i.append(max_outbound_index)\n",
" differenc_set = np.setdiff1d(np.arange(DATABASE_OBJECT_COUNT), fix_i)\n",
" current_db_object_count -= 1"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "11682b36-e705-4bd9-9d75-79012791d1ee",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "graphs",
"language": "python",
"name": "graphs"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.12.7"
}
},
"nbformat": 4,
"nbformat_minor": 5
}

473
01_nb_cncf_optimization/gen_nb_cost_optimization.ipynb Normal file
View File

@@ -0,0 +1,473 @@
{
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"id": "ab5cd7d1-1a57-46fc-8282-dae0a6cc2944",
"metadata": {},
"outputs": [],
"source": [
"import matplotlib.pyplot as plt\n",
"import numpy as np\n",
"import random"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "3d1ad0b9-f6a8-4e98-84aa-6e02e4279954",
"metadata": {},
"outputs": [],
"source": [
"DATABASE_OBJECT_COUNT = 100\n",
"CACHE_SIZE = DATABASE_OBJECT_COUNT/2\n",
"ZIPF_CONSTANT = 2\n",
"\n",
"CACHE_MISS_COST = 2\n",
"CACHE_REFRESH_COST = 1\n",
"\n",
"SEED = 42\n",
"np.random.seed(SEED)\n",
"random.seed(SEED)\n",
"\n",
"LAMBDA_VALUES = np.array([np.random.zipf(ZIPF_CONSTANT) for i in np.arange(1, DATABASE_OBJECT_COUNT + 1,1)])"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "9cc83cf6-5c78-4f0d-b7cb-08cdb80c362e",
"metadata": {},
"outputs": [],
"source": [
"# LAMBDA_VALUES = np.array([0.03, 0.04,0.05,0.06,0.07,1,1.1,1.2,1.3,1.4,1.5])\n",
"# DATABASE_OBJECT_COUNT = len(LAMBDA_VALUES)\n",
"# CACHE_SIZE = 4.4\n",
"# CACHE_MISS_COST = 7\n",
"# CACHE_REFRESH_COST = 1"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "3dc07233-0b56-4fee-a93b-212836c18b42",
"metadata": {},
"outputs": [],
"source": [
"db_object_count = DATABASE_OBJECT_COUNT\n",
"cache_sz = CACHE_SIZE\n",
"\n",
"lambda_vals = LAMBDA_VALUES\n",
"c_f = CACHE_MISS_COST\n",
"c_delta = CACHE_REFRESH_COST"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "5a27d416-8f98-4814-af9e-6c6bef95f4ef",
"metadata": {},
"outputs": [],
"source": [
"def eta_star(db_object_count, c_f, cache_sz, c_delta, lambda_vals):\n",
" num = (db_object_count * c_f - cache_sz * c_delta)\n",
" denom = np.sum(1.0/lambda_vals)\n",
" return max(0, num/denom)"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "6276a9ce-f839-4fe6-90f2-2195cf065fc8",
"metadata": {},
"outputs": [],
"source": [
"def h_i_star(c_f, eta, lambda_vals, c_delta):\n",
" optimized_hitrate = (c_f - (eta/lambda_vals)) / c_delta\n",
" return optimized_hitrate"
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "dcd31a8c-6864-4b9a-8bb3-998f0c32baf6",
"metadata": {},
"outputs": [],
"source": [
"def get_index_of_furthest_hitrate_from_boundary(hitrates):\n",
" local_hitrates = hitrates[(hitrates < 0) | (hitrates > 1)]\n",
" smallest_delta = np.abs(np.min(local_hitrates))\n",
" biggest_delta = np.max(local_hitrates) - 1\n",
" if smallest_delta > biggest_delta:\n",
" index = np.where(hitrates == np.min(local_hitrates))[0][0]\n",
" return index\n",
" else:\n",
" index = np.where(hitrates == np.max(local_hitrates))[0][0]\n",
" return index"
]
},
{
"cell_type": "code",
"execution_count": 18,
"id": "55b251f8-97ca-49a8-9ec6-be77dc1e49b2",
"metadata": {
"scrolled": true
},
"outputs": [],
"source": [
"\"\"\"\n",
"Perform theoretical optimization to compute optimal hit probabilities.\n",
"\n",
"Parameters:\n",
"- lambda_vals (numpy array): Request rates for each item.\n",
"- B (float): Total cache size.\n",
"- c_f (float): Fetching linear cost (cache miss cost).\n",
"- c_delta (float): Age linear cost.\n",
"\n",
"Returns:\n",
"- h_opt (numpy array): Optimal hit probabilities for each item.\n",
"\"\"\"\n",
"optimized_hitrates = np.zeros(DATABASE_OBJECT_COUNT)\n",
"differenc_set = np.arange(DATABASE_OBJECT_COUNT)\n",
"fix_i = []\n",
"current_db_objects = DATABASE_OBJECT_COUNT\n",
"current_cache_size = CACHE_SIZE\n",
"\n",
"while True:\n",
" if current_db_objects == 0:\n",
" # Handle special case: no items left to optimize\n",
" if current_cache_size > 0:\n",
" # Redistribute unused cache size among items with zero hit probability\n",
" differenc_set = np.where(optimized_hitrates == 0)[0]\n",
" fix_i = np.setdiff1d(np.arange(DATABASE_OBJECT_COUNT), differenc_set)\n",
" current_db_objects = len(differenc_set)\n",
" continue\n",
" else:\n",
" optimized_hitrates[differenc_set] = 0\n",
" break\n",
" # Compute Lagrangian multiplier and optimal hit probabilities\n",
" mu = max(0, (current_db_objects * c_f - current_cache_size * c_delta) / np.sum(1.0 / lambda_vals[differenc_set]))\n",
" eta = eta_star(current_db_objects, c_f, current_cache_size, c_delta, lambda_vals[differenc_set])\n",
" assert(mu == eta)\n",
" optimized_hitrates[differenc_set] = (c_f - mu / lambda_vals[differenc_set]) / c_delta\n",
" # print(optimized_hitrates)\n",
" # Handle the case where mu < 0\n",
" if mu < 0:\n",
" current_cache_size = current_db_objects * c_f / c_delta # Adjust cache size for next iteration\n",
" continue\n",
" # Check for constraint violations\n",
" larger_i = np.where(optimized_hitrates > 1)[0] # h > 1\n",
" smaller_i = np.where(optimized_hitrates < 0)[0] # h < 0\n",
" # If no violations, optimization is complete\n",
" break_con = len(smaller_i) == 0 and len(larger_i) == 0\n",
" break_con1 = len((optimized_hitrates[differenc_set])[((optimized_hitrates[differenc_set]) < 0) | ((optimized_hitrates[differenc_set])> 1)]) == 0\n",
" assert(break_con == break_con1)\n",
" if break_con:\n",
" break\n",
" # Find the furthest violating item\n",
" min_viol, min_viol_i = (0, -1)\n",
" if len(smaller_i) > 0:\n",
" min_viol_i = np.argmin(optimized_hitrates)\n",
" min_viol = optimized_hitrates[min_viol_i]\n",
" max_viol, max_viol_i = (0, -1)\n",
" if len(larger_i) > 0:\n",
" larger = optimized_hitrates - 1\n",
" max_viol_i = np.argmax(larger)\n",
" max_viol = larger[max_viol_i]\n",
" # Compare the furthest violations and adjust accordingly\n",
" viol_i = min_viol_i\n",
" min_viol_flag = True # True if furthest is from the left boundary\n",
" if max_viol > abs(min_viol):\n",
" viol_i = max_viol_i\n",
" min_viol_flag = False \n",
" index = get_index_of_furthest_hitrate_from_boundary(optimized_hitrates)\n",
" if viol_i != index:\n",
" print(optimized_hitrates[viol_i])\n",
" print(optimized_hitrates[index])\n",
" assert(viol_i == index)\n",
" if min_viol_flag:\n",
" optimized_hitrates[viol_i] = 0\n",
" else:\n",
" optimized_hitrates[viol_i] = min(1, current_cache_size)\n",
"\n",
" # Update parameters for next iteration\n",
" current_cache_size =- optimized_hitrates[viol_i]\n",
" fix_i.append(viol_i)\n",
" differenc_set = np.setdiff1d(np.arange(DATABASE_OBJECT_COUNT), fix_i)\n",
" current_db_objects = DATABASE_OBJECT_COUNT - len(fix_i)"
]
},
{
"cell_type": "code",
"execution_count": 9,
"id": "efa16eaf-a10b-4927-99cd-190e2ffe1d1e",
"metadata": {},
"outputs": [],
"source": [
"a = optimized_hitrates\n",
"b = differenc_set"
]
},
{
"cell_type": "code",
"execution_count": 10,
"id": "0e21c26f-058a-4e56-a5ad-1c47bf28656c",
"metadata": {
"scrolled": true
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"All values optimized.\n"
]
}
],
"source": [
"\"\"\"\n",
"Perform theoretical optimization to compute optimal hit probabilities.\n",
"\n",
"Parameters:\n",
"- lambda_vals (numpy array): Request rates for each item.\n",
"- B (float): Total cache size.\n",
"- c_f (float): Fetching linear cost (cache miss cost).\n",
"- c_delta (float): Age linear cost.\n",
"\n",
"Returns:\n",
"- h_opt (numpy array): Optimal hit probabilities for each item.\n",
"\"\"\"\n",
"optimized_hitrates = np.zeros(DATABASE_OBJECT_COUNT)\n",
"differenc_set = np.arange(DATABASE_OBJECT_COUNT)\n",
"fix_i = []\n",
"current_db_objects = DATABASE_OBJECT_COUNT\n",
"current_cache_size = CACHE_SIZE\n",
"\n",
"while True:\n",
" if current_db_objects == 0:\n",
" # Handle special case: no items left to optimize\n",
" if current_cache_size > 0:\n",
" # Redistribute unused cache size among items with zero hit probability\n",
" differenc_set = np.where(optimized_hitrates == 0)[0]\n",
" fix_i = np.setdiff1d(np.arange(DATABASE_OBJECT_COUNT), differenc_set)\n",
" current_db_objects = len(differenc_set)\n",
" continue\n",
" else:\n",
" optimized_hitrates[differenc_set] = 0\n",
" break\n",
" # Compute Lagrangian multiplier and optimal hit probabilities\n",
" eta = eta_star(current_db_objects, c_f, current_cache_size, c_delta, lambda_vals[differenc_set])\n",
" optimized_hitrates[differenc_set] = (c_f - eta / lambda_vals[differenc_set]) / c_delta\n",
"\n",
" if mu < 0:\n",
" current_cache_size = current_db_objects * c_f / c_delta # Adjust cache size for next iteration\n",
" continue\n",
" \n",
" if len((optimized_hitrates[differenc_set])[((optimized_hitrates[differenc_set]) < 0) | ((optimized_hitrates[differenc_set])> 1)]) == 0:\n",
" print(\"All values optimized.\")\n",
" break\n",
" max_outbound_index = get_index_of_furthest_hitrate_from_boundary(optimized_hitrates)\n",
" optimized_hitrates[max_outbound_index] = (1 if optimized_hitrates[max_outbound_index] > 1 else 0)\n",
"\n",
" current_cache_size =- optimized_hitrates[max_outbound_index]\n",
" fix_i.append(max_outbound_index)\n",
" differenc_set = np.setdiff1d(np.arange(DATABASE_OBJECT_COUNT), fix_i)\n",
" current_db_objects = DATABASE_OBJECT_COUNT - len(fix_i)"
]
},
{
"cell_type": "code",
"execution_count": 11,
"id": "4f64253f-b389-4be9-b403-08027d480121",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
" 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
" 0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0.,\n",
" 1., 0., 0., 0., 0., 0., 1., 0., 0., 1., 0., 0., 0., 0., 1., 0., 1.,\n",
" 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0.,\n",
" 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0.])"
]
},
"execution_count": 11,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"optimized_hitrates"
]
},
{
"cell_type": "code",
"execution_count": 12,
"id": "17d818db-ec88-4c26-92af-6d74862525d9",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([ 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0.43902439, 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0.43902439, 0. , 0. ,\n",
" 0. , 0.04878049, 0. , 0.04878049, 0. ,\n",
" 0. , 0. , 0. , -0. , 0. ,\n",
" 0.04878049, 0.04878049, 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0.43902439, 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" -0. , 0. , 0. , 0. , 0. ,\n",
" -0. , 0. , 1. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , -0. , 0. ,\n",
" 0. , 0.43902439, 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0. , 0. ,\n",
" 0. , 0. , 0. , 0.04878049, 0. ])"
]
},
"execution_count": 12,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"a"
]
},
{
"cell_type": "code",
"execution_count": 13,
"id": "791b3f96-527a-489e-970e-c92ec950177f",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([ 6, 27, 31, 33, 40, 41, 50, 81, 98])"
]
},
"execution_count": 13,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"b"
]
},
{
"cell_type": "code",
"execution_count": 14,
"id": "c22fa973-432a-4c05-89bf-2a6ea82ae3d2",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([ 6, 27, 50, 81])"
]
},
"execution_count": 14,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"differenc_set"
]
},
{
"cell_type": "code",
"execution_count": 15,
"id": "898e1266-5aaa-46f4-ac0f-c7807ac2b6bb",
"metadata": {},
"outputs": [],
"source": [
"db_object_count = DATABASE_OBJECT_COUNT\n",
"cache_sz = CACHE_SIZE\n",
"loop_lambda = lambda_vals\n",
"\n",
"non_optimized_values = np.arange(db_object_count)"
]
},
{
"cell_type": "code",
"execution_count": 16,
"id": "8cc9b8a9-f7ae-48fc-adfb-ac4b7a4998f1",
"metadata": {},
"outputs": [],
"source": [
"db_object_count = DATABASE_OBJECT_COUNT\n",
"cache_sz = CACHE_SIZE\n",
"loop_lambda = lambda_vals\n",
"\n",
"optimized_hitrate = np.zeros(db_object_count)\n",
"non_optimized_values = np.arange(db_object_count)\n",
"optimized_value = {}\n",
"\n",
"eta = eta_star(db_object_count, c_f, cache_sz, c_delta, loop_lambda[non_optimized_values])\n",
"optimized_hitrate[non_optimized_values] = h_i_star(c_f, eta, loop_lambda[non_optimized_values], c_delta)\n",
"\n",
"max_outbound_index = get_index_of_furthest_hitrate_from_boundary(optimized_hitrate)\n",
"optimized_value[max_outbound_index] = (1 if optimized_hitrate[max_outbound_index] > 1 else 0)"
]
},
{
"cell_type": "code",
"execution_count": 17,
"id": "cbcf3592-fcf2-4f54-a3cd-761097c12972",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"{67: 1}"
]
},
"execution_count": 17,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"optimized_value"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "db732331-1d09-45b7-915c-73daa270b5e2",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "graphs",
"language": "python",
"name": "graphs"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.12.7"
}
},
"nbformat": 4,
"nbformat_minor": 5
}

5
01_nb_cncf_optimization/matlab/Theoritical_opt-Copy1.m
View File

@@ -82,9 +82,6 @@ while flag
differenc_set=setdiff(1:N,fix_i) ;
n=N-length(fix_i);
end
end
end

221
01_nb_cncf_optimization/nb_cost_optimization.ipynb
View File

File diff suppressed because one or more lines are too long

45757
01_nb_cncf_optimization/notes_code.excalidraw Normal file
View File

File diff suppressed because one or more lines are too long

0
01_nb_cncf_optimization/Untitled-2024-11-20-1614.excalidraw → 01_nb_cncf_optimization/notes_paper.excalidraw
View File