Goal

Compare file saving performance of local disk, simple NFS mount, and new NAS with the spex 
instrument software.
Spex can be run in a simulation mode, where it images are read from a disk file. I wish to
compare perforamce of various destination disk. 

Test conditions were not ideal. 
I used the bidgog computer located in my office (on a little 8port switch), and generated and
save data to various computer in the building 

  1-no_saving.txt - just generating data and piping to DV. Autosave OFF.
  2-bigdog.txt    - Autosave ON, to local disk on bigdog /aux partition on it's boot disk.
  3-iarcdata.txt  - Autosave ON, to /netdisks/iarcdata. The iarc lab computer has a 2TB disk on NFS.
  4-dscrs1.txt    - Autosave ON, to /netdisks/dscrs1. This the duke with new NAS hardware.
                    Computer is located in the IRTF student room.

spex was setup with: itime = 0.3, cycles = 2000

With no_saving, the cycle time was about 0.45 seconds. 
The 2000 cycles generated 2000*16MB images, or 32GB.
No save took about 900 seconds (15 mins). So data is generated at 32,000MB/900s or 35.5MB/s.

The data files have 3 columns:
   1. cycle count
	2. time for cycle in seconds.
	3. Total time     in seconds.

Raw data files are here:
	1-no_saving.txt
	2-bigdog_dv.txt
	2-bigdog_nodv.txt
	3-iarcdata_nodv.txt
	4-dscrs1_nodv.txt

DV is athe busies process, seem like it could no keep past with the data viewing.
Even after all the data was done, DV froze, and it process stayed at 100% CPU, for 
a few minutes after the data acquisition was done.
So above I re-ran test 2 with NO dv. I think DV affect cycle time at high data rates.

Graphs

There is a plot of all the data using:
	set yrange [0.2:1.6]
	plot   '1-no_saving.txt' using 1:2 with lines title "1-nosave"
	replot '2-bigdog_dv.txt' using 1:2 with lines title "2-bigdog_dv"
	replot '2-bigdog_nodv.txt' using 1:2 with lines title "2-bigdog_nodv"
	replot '3-iarcdata_nodv.txt' using 1:2 with lines title "3-iarcdata_nodv"
	replot '4-dscrs1_nodv.txt' using 1:2 with lines title "4-dscrs1_nodv"



Using the boot disk to save data is very inconsistant, plus the DV problem make things
worst at after 1200 images.

Removing test #2, here is a plot of no_saving, iarcdata, dscrs1:



hmmm...

Disappointed with the DV issue. There could be some issue with writing, then having
dv read the file. But there seem to be a bug as the CPU percentage continues to be high
every after the cycles are done.

Might be good to put a fast, dedicated /data data in in all the computer. And re-run the local
disk test. I knew it was bad to use the boot disk, but bigdig is a single disk system.

Strange that iarc did better than nduke.