Business resource optimization with Drools Planner

1. Business resource optimization
with
Drools Planner
Geoffrey De Smet

2. Demo
Travelling Salesman Problem
Vehicle Routing
2

3. Every organization has planning problems.
3

4. What is a planning problem?
 Complete goals
 With limited resources
 Under constraints
4

5. Hospital employee rostering
 Assign each
● Shift
 To
● Employee
 Constraints
● Employment contract
http://www.flickr.com/photos/glenpooh/709704564/ ● Free time preferences
● Skills
5

6. Demo
Employee rostering
6

7. Other planning problems
Course scheduling Hospital bed planning
http://www.flickr.com/photos/markhillary/2227726759/
http://www.flickr.com/photos/phelyan/2281095105/
Many more: port planning, airline routing, train composition,
task assignment, assembly line planning,
satellite bandwidth planning, stock cutting, ...
7

8. Bin packing in the cloud
 Assign each
● Process
 To
● Server
 Constraints
● Hardware requirements
● Minimize server cost
http://www.flickr.com/photos/torkildr/3462607995/
8

9. Bin packing in the cloud
 Assign each
● Process
 To
● Server
 Constraints
● Hardware requirements
● Minimize server cost
http://www.flickr.com/photos/torkildr/3462607995/
9

10. Which processes do we assign
to this server?
10

11. How did we find that solution?
11

12. First fit by decreasing size
12

13. First fit by decreasing size
13

14. First fit by decreasing size
14

15. First fit by decreasing size
15

16. Another case
16

17. Try FFD again
17

18. Try FFD again
18

19. Try FFD again
19

20. Try FFD again
20

21. FFD failure
21

22. NP complete
22

23. NP complete
 No silver bullet known
● Holy grail of computer
science
● P == NP
● Probably does not exist
● P != NP
 Root problem of
all planning problems
http://www.flickr.com/photos/annguyenphotography/3267723713/
23

24. Multiple servers...
24

25. Multiple servers...
25

26. Multiple servers...
26

27. Multiple servers...
27

28. Multiple constraints...
28

29. Multiple constraints...
29

30. Multiple constraints...
30

31. Multiple constraints...
31

32. Organizations rarely optimize
planning problems.
32

33. Given 2 solutions, which one is better?
33

34. Each Solution has 1 Score
34

35. Each Solution has 1 Score
35

36. Better score => better solution
36

37. Better score => better solution
37

38. Best score => best solution
38

39. Best score => best solution
39

40. Reuse optimization algorithms?
40

41.  Open source (ASL 2.0)
 Regular releases
 Reference manual
 Examples
41

42. Demo
CloudBalance example
42

43. Domain model

44. Server
44

45. Server
public class Server {
private int cpuPower;
private int memory;
private int networkBandwidth;
private int cost;
// getters
}
45

46. Process
46

47. Process
47

48. Process is a planning entity
@PlanningEntity
public class Process {
private int requiredCpuPower;
private int requiredMemory;
private int requiredNetworkBandwidth;
private Server server;
@PlanningVariable @ValueRange(... = "serverList")
public Server getServer() {...}
// getters, setters, clone, equals, hashcode
}
48

49. CloudBalance
49

50. Solution class CloudBalance
public class CloudBalance
implements Solution<HardAndSoftScore> {
private List<Server> serverList;
private List<Process> processList;
private HardAndSoftScore score;
@PlanningEntityCollectionProperty
public List<Process> getProcessList() {...}
// getters, setters, clone, getProblemFacts()
}
50

51. Score constraints

52. Score calculation
 Simple Java
 Incremental Java
 Drools
52

53. Simple Java score calculation
 Easy to implement
 Bridge to an existing scoring system
 Slow
public class ...
implements SimpleScoreCalculator<CloudBalance> {
public Score calculateScore(CloudBalance cb) {
...
}
}
53

54. Incremental Java
score calculation
 Fast
● Solution changes => only score delta is recalculated
 Hard to implement
● Much boilerplate code
54

55. Drools score calculation
 Also incremental => fast
● but no boilerplate code
 Integration opportunities
● Guvnor repository
● DSL: natural language constraints
● Decision tables
55

56. DRL soft constraint:
server cost
rule "serverCost"
when
// there is a server
$s : Server($c : cost)
// there is a processes on that server
exists Process(server == $s)
then
// lower soft score by $c
end
56

57. DRL hard constraint:
CPU power
rule "requiredCpuPowerTotal"
when
// there is a server
$s : Server($cpu : cpuPower)
// with too little CPU for its processes
$total : Number(intValue > $cpu) from accumulate(
Process(server == $s,
$requiredCpu : requiredCpuPower),
sum($requiredCpu)
)
then
// lower hard score by ($total - $cpu)
end
57

58. Solving it

59. Solver configuration by XML
<solver>
<solutionClass>...CloudBalance</solutionClass>
<planningEntityClass>...Process</>
<scoreDirectorFactory>
<scoreDefinitionType>HARD_AND_SOFT</>
<scoreDrl>...ScoreRules.drl</scoreDrl>
</scoreDirectorFactory>
<!-- optimization algorithms -->
</solver>
59

60. Solving
XmlSolverFactory factory = new XmlSolverFactory(
"...SolverConfig.xml");
Solver solver = factory.buildSolver();
solver.setPlanningProblem(cloudBalance);
solver.solve();
cloudBalance = (CloudBalance) solver.getBestSolution();
60

61. Optimization algorithms

62. Brute Force
62

63. Brute Force config
<solver>
...
<bruteForce />
</solver>
63

64. Brute force scalability
12 processes = 17 minutes
* 680
9 processes = 1.5 seconds
6 processes = 15 ms * 100
64

65. Plan 1200 processes
with brute force?
65

66. First Fit
66

67. First Fit config
<solver>
...
<constructionHeuristic>
<constructionHeuristicType>FIRST_FIT</>
</constructionHeuristic>
</solver>
67

68. First Fit Decreasing
68

69. First Fit Decreasing config
<solver>
...
<constructionHeuristic>
<constructionHeuristicType>FIRST_FIT_DECREASING</>
</constructionHeuristic>
</solver>
@PlanningEntity(difficultyComparatorClass
= ProcessDifficultyComparator.class)
public class Process {
...
}
69

70. First Fit scalability
70

71. Local Search comes after
Construction Heuristics
71

72. Local Search comes after
Construction Heuristics
<solver>
...
<constructionHeuristic>
<constructionHeuristicType>FIRST_FIT_DECREASING</>
</constructionHeuristic>
<localSearch>
...
<localSearch>
</solver>
72

73. Local Search
73

74. Local Search Algorithms
 (Hill climbing)
 Tabu Search
 Simulated Annealing
 Late Acceptance
Details explained Friday 4:15 PM
74

75. Local Search results
75

76. Cost ($) reduction
 Compared to First Fit
● First Fit Decreasing
● 49 480 $ = 4 %
● Tabu Search
● 160 860 $ = 14 %
● Simulated annealing
● 128 950 $ = 11 %
 Few constraints
● => low ROI
76

77. Benchmarker toolkit
77

78. Organizations rarely optimize
planning problems.
78

79. http://www.flickr.com/photos/techbirmingham/345897594/
Organizations waste resources.
79

80. Summary

81. Summary
 Drools Planner optimizes business resources
 Adding constraints is easy and scalable
 Switching/combining algorithms is easy
81

82. Try an example!
82

83. Q&A
 Drools Planner homepage:
● http://www.jboss.org/drools/drools-planner
 Reference manual:
● http://www.jboss.org/drools/documentation
 Download this presentation:
● http://www.jboss.org/drools/presentations
 Twitter: @geoffreydesmet
 Google+: Geoffrey De Smet
83