Closed dwr-psandhu closed 5 years ago
Read a large DSS file with 137 regular time series. Each time series was 15 min interval for 82 years, approx 2.8 million points.
pyhecdss total reading time was 1 min, 3 seconds vs HECDSSVue was 4 min, 50 seconds
Read a large DSS file with 137 regular time series. This time with pydss (part of vtools) using python 2.7 (pydss only works on 64 bit, python 2.7).
pyhecdss total reading time was 1min, 3 seconds vs pydss/vtools was 2 mins, 31 seconds
In summary, pyhecdss is more than twice as fast as pydss/vtools and almost 5 times as fast as HecDSSVue.
My hunch is that there is excessive memory assignment and deletion that was happening with the other two libraries as the underlying functions are all the same.
Closing this issue
Performance test vs DSM2-VISTA pyhecdss total reading time was 1min, 5 seconds vs DSM2-VISTA was 1min, 1 second
In summary, DSM2-VISTA is slightly faster than pyhecdss, which makes it the fastest of them all. I think DSM2-VISTA is as efficient as pyhecdss in memory management and the slight increase in performance is because the JVM is better than the Python VM
This issue is only for catalog and reading and writing time series to DSS files. The timing is done very coarsely and over typical datasets encountered in our work here. it is not meant to be a general test