Introduction
-
What kind of IT resources should be selected?
-
Where should these IT resources be installed (vertical/horizontal dimension)?
-
Which applications should be deployed on which level?
Related work
-
Targeting the factory environment shifting the focus from large amounts of moving clients (users) to fewer but more critical data sources modeling physical as well as logical placement constraints
-
Focusing on the planning and re-planning phase for an IT infrastructure for the green field as well as brownfield applications
-
Solving the NP-hard optimization challenge using MILP (Mixed integer linear programming) in a reasonable amount of time by utilizing the edge level as a constraint for a vertical application and resource placement
Multilayer model for resource placement
Methodology
Model
Sets and parameters
Optimization variables
Constraints
Objective functions
Combining multiple objectives
Implementation in a factory context
Experiment parameters | Repetitions | 100 | ||||
Location sets | 100, 200, 500, 1000 | |||||
App sets | 100, 200, 500, 1000 | |||||
Edge Level | 0 | 1 | 2 | 3 | 4 | |
Distribution (weight) | 0.4 | 0.3 | 0.2 | 0.1 | 0.1 | |
Optimization parameters | Time period | 5 years | ||||
Objective | Minimize cost | |||||
Placement strategy | Upward path | |||||
Optimization configuration | Modeling language | Pyomo | ||||
Solver | SCIP version 7.0.1 | |||||
Termination: Time limit | 10 h | |||||
Termination: Gap limit | 0.01 | |||||
Benchmark PC | CPU | Intel i7-6700 CPU @ 3.40 GHz | ||||
RAM | 36 GB | |||||
OS | Ubuntu 20.04 |
Implementation and benchmark
Implementation
Resources | S | M | L | XL | Cloud |
---|---|---|---|---|---|
Performance unit | 5 | 50 | 500 | 1000 | \(\infty \) |
Memory unit | 1 | 5 | 20 | 70 | \(\infty \) |
Bandwidth limitation | 100 | 100 | 1000 | 1000 | 10,000 |
CapEx (€) | 50 | 500 | 2000 | 7000 | 0 |
OpEx (€/a) | 500 | 1000 | 2000 | 4000 | 0.1 |
Environmental impact (tCO\(_{2}\)/a) | 5 | 15 | 25 | 30 | 0.1 |
Benchmark results
Objectives | |||||
---|---|---|---|---|---|
Initial situation* | Cost | Environmental impact | Number of nodes | ||
Overall optimization time (s) | – | 2.7 | 1.6 | 1.6 | |
Constraint calculation time (s) | – | 0.7 | 0.7 | 0.5 | |
Solve time (s) | – | 1.8 | 0.9 | 1.1 | |
Number of placed resources | 18 | 12 | 12 | 11 | |
Total Cost (€) | 216k | 80k | 80k | 147k | |
Environmental impact (tCO\(_{2}\)) | 457.2 | 135 | 135 | 429.4 | |
\(\hbox {R}_{\textrm{S}}\) | 0 | 4 | 4 | 0 | |
Resource | \(\hbox {R}_{\textrm{M}}\) | 0 | 4 | 4 | 0 |
distribution | \(\hbox {R}_{\textrm{L}}\) | 18 | 4 | 4 | 10 |
\(\hbox {R}_{\textrm{XL}}\) | 0 | 0 | 0 | 1 |