Based on review of the visualization charts of some float data from the below bi dataset. I observed the compression speed, encode time and decode time for some vector sizes could use the alp which is much more superior than the auto selected.
Eg
| name |
data_type |
vector_size |
raw_size |
encoded_size |
num_values |
num_vectors |
bits_per_value |
analyzing_speed(cycles_per_value) |
analyzing_time(ns) |
compression_speed(cycles_per_value) |
compression_time(ns) |
decompression_speed(cycles_per_value) |
decompression_time(ns) |
eq_compare_speed(cycles_per_value) |
eq_compare_time(ns) |
lt_compare_speed(cycles_per_value) |
lt_compare_time(ns) |
rg_compare_speed(cycles_per_value) |
rg_compare_time(ns) |
in_compare_speed(cycles_per_value) |
in_compare_time(ns) |
encoder |
compression_info |
| /auto/CityMaxCapita_1/Created_Date_Time |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1k |
float64 |
1024 |
8192 |
6039 |
1024 |
1 |
47.1797 |
35.8057 |
7333 |
40.4834 |
8291 |
9.1553 |
1875 |
9.7656 |
2000 |
9.3604 |
1917 |
13.8379 |
2834 |
0.2002 |
41 |
alprd |
"ALP-RD(f64)[>>55][Dict(u16)[Raw(u16)[n=1]][Const(u16)[n=1024]]][BP(u64)[w=47 |
| /fixed/CityMaxCapita_1/Created_Date_Time/alp |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1k |
float64 |
1024 |
8192 |
17680 |
1024 |
1 |
138.1250 |
21.9727 |
4500 |
50.8594 |
10416 |
18.5107 |
3791 |
4.4775 |
917 |
24.2139 |
4959 |
19.7314 |
4041 |
0.2051 |
42 |
alp |
"ALP(f64)_[0 |
Data Used from BI
- CityMaxCapita_1, Created_Date_Time
- SalariesFrance_1, AG_25_29
Based on review of the visualization charts of some float data from the below bi dataset. I observed the compression speed, encode time and decode time for some vector sizes could use the alp which is much more superior than the auto selected.
Eg
Data Used from BI