5171 Rate this article:
4.0

Does Solid State Drive (SSD) storage provide benefits for ENVI and IDL?

Adam O'Connor
First and foremost, the title of this blog post is somewhat of a rhetorical question as we've all heard about the benefits of Solid State Drive (SSD) storage by now. I recently decided that my seven year old desktop computer at home desperately needed an upgrade (it still had Windows Vista!) and purchased a new workstation. During the buying process research, I was presented with a number of hardware configuration "upgrade" options ranging from 6+ core processors, high-end graphics cards, additional memory, and Blu-Ray disc drive. In the end, I decided that I will get the most bang for my buck by selecting the option to have a Solid State Drive (SSD) storage device as the boot hard drive. I chose a 256GB SSD so there is enough room for the operating system, software installations and critical in-use files with a secondary Hard Disk Drive (HDD) for permanent archival storage.
 
Immediately upon starting the new workstation I noticed numerous performance benefits over my old computer (as expected). The time it takes to boot up the machine is greatly reduced, software applications launch almost instantaneously, and working with files on the SSD was extremely fast. Certainly a lot of this can be attributed to the benefits of Windows 7 over Vista, faster processor, additional memory, modern graphics card, etc. But, I started to wonder just how much of the performance can be attributed to the extra money I spent on the SSD storage device? In particular, are there significant performance benefits for our IDL and ENVI software when performing fairly common operations that happen to be "disk intensive" in nature?
 
The only way to acquire accurate performance benchmarks is to have two computers with the same exact hardware then modify one variable (in this case the boot drive) but I don't have these resources at my disposal. Consequently, I choose to use a machine that has the closest configuration match to the new workstation. At the bottom of this post you will find some of the more relevant specifications for the two computers used in my testing exercise. It is also worth mentioning that all tests were performed with the computers in a freshly rebooted state using standard ENVI raster format (i.e. simple binary) files with BSQ interleave stored directly on the C: boot drive (with processing output results written to same drive).
 
The following performance tests I performed are by no means comprehensive and do not represent an official Exelis Visual Information Solutions stance on the value of Solid State Drive (SSD) storage. This testing exercise was merely to satisfy my own curiosity in the very limited time (a few hours) I had to devote to such an activity and I share the results with you here in a casual fashion. In other words, this is not intended to be interpreted as accurate performance benchmarks but the results are interesting nonetheless. So without further ado …
 
IDL - Launch IDL 8.4 (Workbench):
   Computer with HDD = 0:42
   Computer with SSD = 0:06
 
IDL - Read 1GB Binary Data into Variable (READU):
   Computer with HDD = 0:16
   Computer with SSD = 0:03
 
ENVI - Launch ENVI 5.2:
   Computer with HDD = 0:27
   Computer with SSD = 0:07
 
ENVI - Build Pyramid for 2GB PAN Raster:
   Computer with HDD = 1:36
   Computer with SSD = 0:17
 
ENVI - Compute Segmentation for 64MB PAN Raster:
   Computer with HDD = 2:46
   Computer with SSD = 1:47
 
ENVI - Compute SAVI Spectral Index for 2GB MSI Raster:
   Computer with HDD = 0:36
   Computer with SSD = 0:05
 
ENVI - Compute K-Means Classification for 2GB MSI Raster:
   Computer with HDD = 5:06
   Computer with SSD = 1:12
 
ENVI - Compute MNF for 64MB HSI Raster:
   Computer with HDD = 1:18
   Computer with SSD = 0:12
 
Considering there are a wide variety of hardware configuration differences between the two computers (see details below) it is quite easy to poke holes in the methodology used in this benchmarking exercise, but from my perspective the performance data at least supports the hypothesis that Solid State Drive (SSD) storage provides significant benefits for IDL and ENVI. I now feel confident that my decision to purchase a SSD boot drive was money well spent which is, of course, the primary impetus for this testing exercise.
 
Computer with HDD
  • Operating System: Windows 7 Professional 64-bit Service Pack 1
  • Processor: Intel Core i7-2620M CPU @ 2.70GHz
  • Number of Cores: 2
  • Number of Logical Processors: 4
  • Memory: 4096MB 1600MHz DDR3 RAM
  • Video Card: NVIDIA NVS 4200M 
  • Video Memory: 512MB
  • Boot Drive: 3.5" Serial ATA (7,200 Rpm) Hard Disk Drive
Computer with SSD
  • Operating System: Windows 7 Professional 64-bit Service Pack 1
  • Processor: Intel Core i7-4790 CPU @ 3.60GHz
  • Number of Cores: 4
  • Number of Logical Processors: 8
  • Memory: 16384MB 1600MHz DDR3 RAM
  • Video Card: NVIDIA Quadro K600
  • Video Memory: 1024MB
  • Boot Drive: 2.5" Serial ATA Solid State Drive

4 comments on article "Does Solid State Drive (SSD) storage provide benefits for ENVI and IDL?"

Michael Galloy

Since your new system has both a SSD and HDD, couldn't you perform the both tests with that single system (except maybe for the pain of installing ENVI/IDL on both and timing the launch times)? The SSD system above is quite a bit beefier than the HDD system.


Andrew Cool

I run a 3930k, 6 core system with a separate smaller SSD for Windows 7 OS to pummel, another larger SSD for intensive data access, and then a couple of archival HDDs. The systems sings for the 32000 PNG files that I generate daily for my IDL driven astronomy weather website.

Can thoroughly recommend taking old plate mechanical spinners out of the line of fire!

Andrew Cool

www.skippysky.com.au


Adam O'Connor

Yes, application launch time is one of the key elements I was hoping to compare and cannot install the same version of ENVI/IDL twice on the same machine (and don’t want to compare different versions) but the processing can be benchmarked using the same computer’s HDD storage. I added another pyramid creation plus a benchmark of NNDiffuse Pan-Sharpening which clearly illustrates that processing time all depends on the nature of the algorithm. In the case of NNDiffuse and Segmentation the key performance factor is clearly not disk I/O. In contrast, READU, pyramid creation, unsupervised classification and MNF all clearly benefit from SSD storage:

IDL - Read 1GB Binary Data into Variable (READU):

Data on HDD = 0:07

Data on SSD = 0:03

ENVI - Build Pyramid for 2GB PAN Raster:

Data on HDD = 0:28

Data on SSD = 0:17

ENVI - Build Pyramid for 2GB MSI Raster:

Data on HDD = 0:21

Data on SSD = 0:13

ENVI - Compute NNDiffuse for 1.2GB PAN+MSI Raster:

Data on HDD = 10:54

Data on SSD = 11:03

ENVI - Compute Segmentation for 64MB PAN Raster:

Data on HDD = 1:54

Data on SSD = 1:47

ENVI - Compute SAVI Spectral Index for 2GB MSI Raster:

Data on HDD = 0:07

Data on SSD = 0:05

ENVI - Compute K-Means Classification for 2GB MSI Raster:

Data on HDD = 2:25

Data on SSD = 1:12

ENVI - Compute MNF for 64MB HSI Raster:

Data on HDD = 1:52

Data on SSD = 0:12


shaun heatheridge

I have a used system (http://www.spectra.com/EMC/used-system/171/index.htm) and inegration of of files in the storage isn't an issue using this kind of parameters. Love your detailed info here.

Please login or register to post comments.