yth98 / ADA2020-Challenge

Team Challenge on Parallel Scheduling Problem. (Algorithm Design and Analysis, Autumn 2020, NTU CSIE)

ADA Final Challenge 2020 - Team31

Usage

• ./scheduler 01.in 01.out

Build and Run-time Dependency

• GCC 7.5+
• Google OR-Tools
• /usr/include/ortools, /usr/lib/libortools.so
• make libs-or/lib/libortools.so would download and build OR-Tools automatically
• make scheduler if OR-Tools is installed

TODO

• Generate valid outputs
• Translate testcases to MIP CP problem instances
  • Solvers feasible with APIs
    • OR-Tools - CP-SAT
    • SCIP
    • GLPK
    • lp_solve
    • Comparison of Open-Source Linear Programming Solvers
    • Mixed Integer Programming Models for Job Shop Scheduling: A Computational Analysis
  • Illustrate the objective function
    • The metric to be minimized
  • Illustrate the constraints: Disjunctive Model
    • Dependencies of operations
    • Capacity of each slice
• Find suitable FPTAS
• Related researches
  • Parallel task scheduling problem - Wikipedia
  • Scheduling for Parallel Processing
    • parallel identical processors
    • the number of processors is a variable
    • nonpreemptive
    • task graph is arbitrary
    • processing times are upper- and lower-bounded by 1 and 96
    • all tasks are ready at time 0
    • no task has any deadline
    • optimality criterion:
      • schedule length (C_max)
      • weighted sum of completion times (\sum {w_j c_j})
    • rigid, noncontiguous
  • Scheduling multiprocessor tasks — An overview
  • Scheduling Parallel Tasks: Approximation Algorithms
  • Bounds on Multiprocessing Timing Anomalies
  • Approximation Algorithms for Scheduling Parallel Jobs
  • Scheduling Multiprocessor Tasks to Minimize Schedule Length - independent, unit length
• Reduce the execution time
  • Multithread / Parallel programming
    • SatParameters::set_num_search_workers
    • SCIPsolveConcurrent (may be slower on Workstations)
  • Divide & Conquer
    • Split jobs to groups
  • Branch pruning
  • Set the time limit explicitly
    • SatParameters::set_max_time_in_seconds(...)
    • SCIPsetRealParam(..., "limits/time", ...)
    • The sub-optimal solution may be worse than traditional one
    • Put the most effort in Testcase 10

Public

#	Slices	Ops	jjl	Our	Strong	Remark
0	3	5	75	51	51	Opt
1	1	172	29584	67900.09074	67900.09074	Opt
2	3	11	363	173.977	173.977	Opt
3	2	14	392	109757.124	109757.124	Opt
4	2	17	578	270386.62151	270533.85177	Opt
5	4	34	4624	12667.7143	12667.7143	Opt
6	4	33	4356	19501.14172	19501.36364	Opt
7	4	28	3136	17533.5815	17695.212	Opt
8	8	40	12800	43642.98051	45288.91921	Strong
9	8	87	60552	64573.2938	68919.2094	Strong
10	36	241	2090916	389586.49585	415366.00282	Strong

Private

#	l	n	Ops	Metric	Remark
c031	1	1	1	1	Optimal
3d2a	1	17	100	677320.389817	Optimal
5626	2	1	1	388.47968	0s Optimal
c8c5	2	30	100	2094950.036765	120s CP-SAT
4ee8	2	23	100	683412.200993	60s CP-SAT
2e58	3	3	6	214	0s Optimal
b06e	3	5	11	101.77677	0s Optimal
1f21	3	7	26	61206.753544	60s CP-SAT
187b	3	6	26	5553	50s Optimal
240e	3	6	26	23769	120s Optimal
1955	3	20	60	8076.669	360s CP-SAT
c53c	3	17	60	16099	360s CP-SAT
4b3b	3	20	65	13167.5427	360s CP-SAT
21ca	3	30	100	1452587.795588	360s CP-SAT
6e33	3	10	100	7662	360s CP-SAT
01ef	4	5	18	564.8458	*Our 180s CP
d311	4	6	25	845.1861	*Our 180s CP
657e	4	15	34	14960.48323	480s CP-SAT
281e	4	3	77	158807.638315	480s CP-SAT
fc0f	4	4	100	386012.897789	480s CP-SAT
a72a	5	2	7	6708	0s Optimal
f29f	5	11	17	31237.063741	600s CP-SAT
368c	5	4	21	4638	100s Opt-MIP
e238	5	10	38	153009.650022	600s CP-SAT
af14	5	25	99	43638.2	720s CP-SAT
1444	5	25	100	507194.93133	720s CP-SAT
3e1e	5	5	100	149457.279682	720s CP-SAT
bc5f	6	6	37	86108.7	180s CP-SAT
1516	6	10	40	7610.389	Traditional
64d1	6	2	47	127042.557474	900s CP-SAT
ddf1	6	15	78	238856.693142	1200s CP-SAT
080c	6	11	83	105800.321041	1200s CP-SAT
1ff9	6	16	100	*876060.34	*Traditional
7118	6	10	100	2083949.872824	1200s CP-SAT
e9eb	6	10	100	2363918.966663	900s CP-SAT
5247	7	6	42	48515.092067	1080s CP-SAT
778c	7	10	50	149754.138889	1080s CP-SAT
4467	7	19	67	33029.89308	Traditional
4551	7	27	76	36845.17234	Traditional
210c	7	12	83	269026.028198	1080s CP-SAT
7d9b	7	22	85	725300.748775	Traditional
621f	8	4	33	570	1200s CP-SAT
ec0d	8	15	40	61303.06067	1200s CP-SAT
95ca	8	30	56	787465.26548	1200s CP-SAT
1b3d	8	29	73	1242802.600223	Traditional
7427	8	10	73	758.102972	1200s CP-SAT
60a9	8	19	75	576.092008	1200s CP-SAT
94a0	8	10	84	388724.507176	1200s CP-SAT
20e3	8	25	85	880111.653036	1200s CP-SAT
d6af	8	8	87	308061.409469	1200s CP-SAT
abaf	8	10	88	291132.838896	1200s CP-SAT
3419	8	30	89	943287.678979	1200s CP-SAT
5204	8	30	89	346093.05308	Traditional
8a44	8	30	90	45648.958668	Traditional
fd5b	8	30	90	386142.42016	Traditional
cc41	8	29	91	312216.35186	Traditional
18e7	8	17	93	2132	Traditional
2542	8	30	94	779116.665037	Traditional
10e2	8	26	96	1147528.548496	Traditional
59bf	8	20	96	2441	Traditional
6407	8	30	97	456539.268123	Traditional
b41d	8	30	97	112697	Traditional
d652	8	30	97	302854	Traditional
2c5f	8	30	100	834279.616973	Traditional
46ad	8	30	100	1477416.14074	Traditional
5905	8	30	100	1208250.244	CP-SAT
59a5	8	30	100	116.534	Traditional
9052	8	30	100	1192972.674	CP-SAT
fb0f	8	30	100	1513040.322707	CP-SAT
3672	8	25	100	1335276.067131	Traditional
b779	8	22	100	1059680.409904	Traditional
cda6	8	19	100	1858	Traditional
96fa	8	18	100	663633	Traditional
4b93	8	12	100	9276.196777	Traditional
d829	8	7	100	155595.28125	Traditional
0942	8	5	100	495290.452062	Traditional
7725	8	3	100	151608	Traditional
77a2	8	1	100	20485.359342	Traditional