PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
December 2020
723
GIS
&
Tips Tricks
By Dave Maune, Ph.D., CP, GS, PS,
and
Al Karlin, Ph.D, CMS-L, GISP
Default settings are not always fast
Conventional wisdom recommends using the fastest media
possible while performing GIS computations to optimize
computing time. Reading/writing to a local physical disk-
drive is generally considered to be faster and more efficient
than using network storage. However, in a production envi-
ronment, where networked servers are used, the temporary
files are frequently, by default, directed to network storage.
While this tip focuses on a specialized piece of GIS-pro-
cessing software, RiPRECISION, it points to the universal
reliability of accepting the “defaults” and that we should
perform our own evaluations.
RiPRECISION is a
RIEGL
additional licensed module with-
in the RiPROCESS software suite. The RiPROCESS Suite
is designed for managing, processing, analyzing and visu-
alizing data acquired with multiple types of laser scanning
sensors. RiPRECISION is used for;
y
adjusting the GNSS/INS trajectory of aircraft (UAVs,
terrestrial vehicles, and all kinematic systems),
y
merging scan data together by adjusting the trajectories
based on matching planer surfaces, accuracy models,
error models and user provided control points and sur-
faces merging overlapping LiDAR scan data, and
y
adjusting data to fit control points.
The time required to perform these adjustments can
increase significantly if the project is stored in a way that
restricts computer I/O operations, especially in a networked
processing environment. The first step in RiPRECISION is
to generate planar surfaces for all laser scan data. As this
process step requires disk space and time, it can be sped up
by adjusting the storage location.
When using RiPRECISION to merge laser scan data and to
adjust the data to control points, it operates from temporary
files rather than the raw data source. The process can be
sped up by moving the temporary file to whatever storage
is faster relative to wherever the raw data are stored. The
temporary files are much smaller than the overall project
folder, so size is less of an issue. This trick works particu-
larly well if a project is stored on a network drive and the
temporary files are moved to a local solid-state drive (SSD)
on a workstation. For example, a 10 TB project generally
will produce temporary files of around 800 Gb. Moving that
temporary files off the network to a local HDD reduced the
processing time from more than a week to a just few days.
Of course, network infrastructure, speed and usage also fac-
tor into the total computation time. Table 1 shows theoret-
ical (manufacturer specifications) transfer speeds of several
commonly used devices. From Table 1, it is obvious that any
local device has a faster transfer speed than 100 MegaBit
(MB) Ethernet and that some newer SSD devices may be
faster than 1000MB Ethernet.
Table 1.
Device
Transfer Speed
(as per specs)
High-speed Ethernet
100 MB/sec
Western Digital Red 4TB HDD
150 MB/sec
Western Digital Ultrastar DC HC300 10TB HDD 250 MB/sec
USB 3
500 MB/sec
Samsung 860 EVO M.2 1TB SSD
520 MB/sec
SATA
600 MB/sec
Samsung 970 EVO Plus M.2 1TB SSD
3300 MB/sec
With RiPROCESS, it is relatively simple to change the
default storage path for temporary files. On the main menu,
click on the Tool dropdown, and select the “Options…”.
Jackson Beebe and Al Karlin, Ph.D.,
CMS-L, GISP
The process can be sped up
by moving the temporary file
to whatever storage is faster
relative to wherever the raw
data are stored.
Photogrammetric Engineering & Remote Sensing
Vol. 86, No. 12, December 2020, pp. 723–724.
0099-1112/20/723–724
© 2020 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.86.12.723
