The following tables shows TPCH results on a single NUC computer using different single node and 10 node cluster configurations.

A node is an Intel NUC i7-8650U, 4 CPU's with 32GB RAM.

1 TPCH Query results

Before going into deep, retain few numbers:

  • lineitem at SF=100 occupies 64GB. This is about 4.2M 16K pages. Informix configuration uses 1.2M 16K buffers (18GB). This means a sequential scan on lineitem will never fit in memory as it's +3 times available buffer size.
  • In contrast a SF=10 lineitem occupies about 7GB, so it clearly fit's in memory.

1.1 Query times

The following table shows query times in seconds for different scale factors for Informix with degree of parellelism 0 and 100 and Hadoop Hive using Map Reduce (additional test should be done with Tez and Spark).

Scales from 1 to 100 are done in single node configurations. Scale 1000 is done in a 10 node configuration.

The following versions have been used:

  • Informix 12.10.FC1 (PDQ 0:100)
  • Hadoop 3.0.3
  • Hive 3.1.1

Table shows time scales from green to red. Red upper limit is set to 10 minutes (600 seconds)

Nodes 1 10
Query SF=1 SF=10 SF=20 SF=50 SF=100 SF=1000
PDQ
0 		PDQ
100 		PDQ
0 		PDQ
100 		Hive PDQ
0 		PDQ
100 		Hive PDQ
0 		PDQ
100 		Hive PDQ
0 		PDQ
100 		Hive PDQ
100 		Hive
Q1 10 4 103 32 162 202 68 295 498 169 710 1005 340 1401 350 .
Q2 0 0 15 4 226 48 30 262 89 62 371 191 145 690 . .
Q3 1 0 46 15 273 189 72 488 531 192 1119 1079 248 2203 . .
Q4 0 1 6 9 237 29 22 425 383 362 982 771 222 1928 . .
Q5 0 0 8 5 169 32 29 215 387 74 363 801 254 HANGS (5h+) . .
Q6 1 1 9 13 90 21 20 164 42 31 390 68 62 764 79 .
Q7 2 0 37 20 225 165 43 474 334 52 368 694 237 2653 . .
Q8 0 1 15 17 391 153 29 629 498 67 1374 1006 321 KILLED (6h+) . .
Q9 2 1 105 36 330 265 230 557 802 570 2152 1958 1073 KILLED (6h+) . .
Q10 1 1 32 4 258 89 56 436 471 156 473 966 247 1990 . .
Q11 0 0 4 1 165 31 11 180 73 35 224 151 49 383 42 .
Q12 1 1 19 8 243 89 14 432 180 49 1001 445 76 1955 110 .
Q13 5 5 54 55 131 114 116 178 304 320 364 723 1214 621 . .
Q14 1 0 11 7 176 15 17 315 86 24 745 170 47 1459 81 .
Q15 1 1 23 10 352 35 18 619 92 50 1415 138 104 2743 . .
Q16 1 1 4 2 177 14 4 226 35 10 361 80 20 527 29 .
Q17 0 0 2 1 434 9 14 791 61 66 1868 123 86 3730 145 .
Q18 3 1 34 17 491 76 37 887 187 140 2112 373 329 4135 629 .
Q19 9 2 9 9 181 24 18 330 57 42 781 226 936 1529 102 .
Q20 0 1 6 10 182 20 87 247 368 560 452 945 694 1168 . .
Q21 18 27 184 267 737 374 500 1345 927 1337 3335 1866 2203 6399 . .
Q22 1 2 12 20 144 24 42 192 58 116 328 114 268 560 . .

1.2 Query times CPU Benchmark

Next table shows query times in seconds for TPCH Scale 1 executed in Informix Server in multiple CPU types. This tables compares CPU power of computation.

The Informix version used is Informix 12.10.FC1

CPUs i7-8650U E5-2680 - 6 CPUVP
Query PDQ
0		 PDQ
100		 PDQ
0		 PDQ
100
Q1 10 3 21 7
Q2 2 1 0 0
Q3 13 0 24 3
Q4 1 1 5 0
Q5 1 0 9 2
Q6 1 1 2 1
Q7 1 0 36 9
Q8 1 1 22 4
Q9 4 1 120 36
Q10 1 0 7 1
Q11 0 0 0 1
Q12 1 1 2 1
Q13 5 5 10 11
Q14 2 0 1 1
Q15 2 1 22 2
Q16 0 0 1 1
Q17 0 0 0 0
Q18 4 1 8 3
Q19 1 1 14 6
Q20 0 1 1 1
Q21 11 19 34 58
Q22 0 1 0 2

2 TPCH Query tables

Query Tables affected
Q1 lineitem
Q2 part, supplier, partsupp, nation, region
Q3 customer, orders, lineitem
Q4 orders, subquery(lineitem)
Q5 customer, orders, lineitem, supplier, nation, region
Q6 lineitem
Q7 subquery (supplier, lineitem, orders, customer, nation n1, nation n2)
Q8 subquery (part, supplier, lineitem, orders, customer, nation n1, nation n2, region)
Q9 subquery (part, supplier, lineitem, partsupp, orders, nation)
Q10 customer, orders, lineitem, nation
Q11 partsupp, supplier, nation
Q12 orders, lineitem
Q13 subquery (customer, orders)
Q14 lineitem, part
Q16 partsupp, part
Q17 lineitem, part
Q18 customer, orders, lineitem
Q19 lineitem, part
Q20 supplier, nation, subquery (partsupp)