- work in progress.
NOTE: All evaluation include the load time and save time on HDFS/Hbase.
Hama 0.19
Matrix Multiplication
Node | Size | Density | Blocking Job | Multiplication |
5 | 5,000 * 5,000 | 100% | 21mins, 32sec | 13mins, 17sec |
Node | Size | Density | Blocking Job | Multiplication |
10 | 5,000 * 5,000 | 100% | 13mins, 31sec | 5mins, 25sec |
Matrix Transpose
Node | Size | Density | Running for |
5 | 5,000 * 5,000 | 100% | 19mins, 48sec |
Node | Size | Density | Running for |
10 | 5,000 * 5,000 | 100% | 12mins, 24sec |
Matrix Norm
Node | Type | Size | Density | Running for |
5 | One | 5,000 * 5,000 | 100% | 9mins, 32sec |
5 | Infinity | 5,000 * 5,000 | 100% | 8mins, 11sec |
5 | Maxvalue | 5,000 * 5,000 | 100% | 8mins, 58sec |
5 | Frobenius | 5,000 * 5,000 | 100% | 8mins, 42sec |
Node | Type | Size | Density | Running for |
10 | One | 5,000 * 5,000 | 100% | 2mins, 35sec |
10 | Infinity | 5,000 * 5,000 | 100% | 2mins, 19sec |
10 | Maxvalue | 5,000 * 5,000 | 100% | 2mins, 22sec |
10 | Frobenius | 5,000 * 5,000 | 100% | 2mins, 25sec |
Benchmarks
This performance contains data load and export operations.
Dependencies Information :
- Hadoop 0.18.2
- Hbase 0.18.1
Hardware Information :
- 4 Intel(R) Xeon(R) CPU 2.33GHz, SATA hard disk, Physical Memory 16,626,844 KB
- Dense matrix add
- Dense matrix multiply
NOTE that 10,000 by 10,000 matrix takes 800MB and 1 hour on single node.
Version | Operation | Cluster Size | Rows | Columns | Total Maps | Total Reduces | Time (seconds) | Bytes Written |
Trunk 712655 | Add | 2 node | 1,000 | 1,000 | 2 | 2 | 17 seconds | 66,326,104 |
Trunk 712658 | Mult | 2 node | 300 | 300 | 2 | 2 | 181 seconds | 5,929,512 |
Version | Operation | Cluster Size | Rows | Columns | Total Maps | Total Reduces | Time (seconds) | Bytes Read | Bytes Written |
Trunk 718158 | Mult | 2 node | 300 | 300 | 2 | 2 | 12 seconds | 1,464,484 | 2,929,092 |
Trunk 720735 | Mult | 2 node | 1,000 | 1,000 | 2 | 2 | 20 seconds | 16,166,452 | 32,333,028 |
Trunk 722320 | Mult | 2 node | 3,000 | 3,000 | 4 | 2 | 124 seconds | 590,672,392 | 872,228,808 |
| No Format |
|---|
NOTE: The following numbers are obtained by using poe+ on the entire code, including minimal I/O and matrix construction.
Matrix-Matrix Multiply of 5,000 by 5,000 dense matrix
Mflip/s Wall sec Library
------- -------- -------------------------------------------
8,300 30 PESSL PDGEMM (16 processors)
7,900 32 ScaLAPACK routine PDGEMM (16 processors)
7,900 32 ESSL-SMP routine DGEMM (16 threads)
7,900 32 NAG-SMP routine F01CKF (16 threads)
1,200 213 ESSL routine DGEMM
Matrix-Matrix Multiply of 20,000 by 20,000 dense matrix
Mflip/s Wall sec Library and configuration
------- -------- -------------------------------------------
158,900 100 ScaLAPACK PDGEMM (256 proc, 16 nodes)
146,200 110 PESSL PDGEMM (256 proc, 16 nodes)
105,400 150 ScaLAPACK PDGEMM (144 proc, 9 nodes, block 128)
100,960 160 PESSL PDGEMM (144 proc, 9 nodes, block 128)
79,400 200 PESSL PDGEMM (144 proc, 9 nodes, block 1024)
74,800 214 ScaLAPACK PDGEMM (144 proc, 9 nodes, block 1024)
55,000 290 PESSL PDGEMM (64 proc, 4 nodes)
50,000 320 ScaLAPACK PDGEMM (64 proc, 4 nodes)
27,160 590 PESSL PDGEMM (32 proc, 2 nodes)
25,630 625 ScaLAPACK PDGEMM (32 proc, 2 nodes)
15,800 1,010 PESSL PDGEMM (16 Proc, 1 node)
15,600 1,025 ScaLAPACK PDGEMM (16 Proc, 1 node)
Matrix-Matrix Multiply of Larger Dense Matrix
Gflip/s Wall sec Size Library and configuration
------- -------- ------- -------------------------------------------
163.6 1,529 50,000 ScaLAPACK PDGEMM (256 proc, 16 nodes)
163.4 1,531 50,000 PESSL PDGEMM (256 proc, 16 nodes)
179.6 11,141 100,000 PESSL PDGEMM (256 proc, 16 nodes, 128 block)
210.7 9,495 100,000 ScaLAPACK PDGEMM (256 proc, 16 nodes, 128 block)
|
...