NERSCPowering Scientific Discovery for 50 Years

2005 User Survey Results

Software

 

  • Legend
  • Software Satisfaction - by Score
  • Software Satisfaction - by Platform
  • Software Comments

 

 

Legend:

SatisfactionAverage Score
Very Satisfied 6.50 - 7.00
Mostly Satisfied 5.50 - 6.49
Somewhat Satisfied 4.50 - 5.49
Significance of Change
significant increase
not significant

 

Software Satisfaction - by Score

7=Very satisfied, 6=Mostly satisfied, 5=Somewhat satisfied, 4=Neutral, 3=Somewhat dissatisfied, 2=Mostly dissatisfied, 1=Very dissatisfied

ItemNum who rated this item as:Total ResponsesAverage ScoreStd. Dev.Change from 2004
1234567
PDSF SW: C/C++ compilers         1 9 18 28 6.61 0.57 0.37
GRID: Job Submission     1   3 9 27 40 6.53 0.85  
GRID: Job Monitoring     1 1 1 11 26 40 6.50 0.88  
SP SW: Fortran compilers     2 2 3 33 65 105 6.50 0.81 0.08
PDSF SW: Programming libraries         1 10 11 22 6.45 0.60 0.32
PDSF SW: Software environment       1   14 15 30 6.43 0.68 0.08
GRID: Access and Authentication     2 2 2 7 30 43 6.42 1.10  
SP SW: Programming libraries     1 6 1 35 57 100 6.41 0.87 0.15
SP SW: Software environment   1   3 3 54 60 121 6.39 0.78 0.05
SP SW: C/C++ compilers   1 1 4 2 25 46 79 6.37 1.00 0.11
GRID: File Transfer     2 2 3 11 25 43 6.28 1.10  
PDSF SW: Fortran compilers     1 1   5 8 15 6.20 1.21 0.33
PDSF SW: General tools and utilities       1 2 13 9 25 6.20 0.76 0.37
SP SW: Applications software     3 2 3 32 27 67 6.16 0.98 0.03
Jacquard SW: C/C++ compilers 1     3 1 24 19 48 6.15 1.09  
PDSF SW: Applications software       2 1 11 8 22 6.14 0.89 0.35
Jacquard SW: Software environment 1   1 1 7 32 25 67 6.12 1.02  
SP SW: General tools and utilities     4 3 8 37 34 86 6.09 1.02 0.18
SP SW: Performance and debugging tools 1   3 3 10 40 30 87 6.00 1.10 0.16
PDSF SW: Performance and debugging tools       2 3 11 7 23 6.00 0.90 0.23
Jacquard SW: General tools and utilities       7 2 20 15 44 5.98 1.02  
Jacquard SW: Programming libraries 1   4 3 2 22 21 53 5.92 1.36  
Jacquard SW: Applications software 1 1 1 3 3 14 13 36 5.78 1.48  
Jacquard SW: Fortran compilers 1 2 1 3 8 20 16 51 5.73 1.40  
SP SW: Visualization software 1 1 2 4 2 18 12 41 5.67 1.47 0.27
Jacquard SW: Visualization software       4 1 8 2 15 5.53 1.06  
DaVinci SW: Visualization software 1   1 1 2 5 4 14 5.43 1.74  
Jacquard SW: Performance and debugging tools 1   4 4 6 15 7 37 5.35 1.44  

 

Software Satisfaction - by Platform

7=Very satisfied, 6=Mostly satisfied, 5=Somewhat satisfied, 4=Neutral, 3=Somewhat dissatisfied, 2=Mostly dissatisfied, 1=Very dissatisfied

ItemNum who rated this item as:Total ResponsesAverage ScoreStd. Dev.Change from 2004
1234567
DaVinci SW: Visualization software 1   1 1 2 5 4 14 5.43 1.74  
GRID: Job Submission     1   3 9 27 40 6.53 0.85  
GRID: Job Monitoring     1 1 1 11 26 40 6.50 0.88  
GRID: Access and Authentication     2 2 2 7 30 43 6.42 1.10  
GRID: File Transfer     2 2 3 11 25 43 6.28 1.10  
Jacquard SW: C/C++ compilers 1     3 1 24 19 48 6.15 1.09  
Jacquard SW: Software environment 1   1 1 7 32 25 67 6.12 1.02  
Jacquard SW: General tools and utilities       7 2 20 15 44 5.98 1.02  
Jacquard SW: Programming libraries 1   4 3 2 22 21 53 5.92 1.36  
Jacquard SW: Applications software 1 1 1 3 3 14 13 36 5.78 1.48  
Jacquard SW: Fortran compilers 1 2 1 3 8 20 16 51 5.73 1.40  
Jacquard SW: Visualization software       4 1 8 2 15 5.53 1.06  
Jacquard SW: Performance and debugging tools 1   4 4 6 15 7 37 5.35 1.44  
PDSF SW: C/C++ compilers         1 9 18 28 6.61 0.57 0.37
PDSF SW: Programming libraries         1 10 11 22 6.45 0.60 0.32
PDSF SW: Software environment       1   14 15 30 6.43 0.68 0.08
PDSF SW: Fortran compilers     1 1   5 8 15 6.20 1.21 0.33
PDSF SW: General tools and utilities       1 2 13 9 25 6.20 0.76 0.37
PDSF SW: Applications software       2 1 11 8 22 6.14 0.89 0.35
PDSF SW: Performance and debugging tools       2 3 11 7 23 6.00 0.90 0.23
SP SW: Fortran compilers     2 2 3 33 65 105 6.50 0.81 0.08
SP SW: Programming libraries     1 6 1 35 57 100 6.41 0.87 0.15
SP SW: Software environment   1   3 3 54 60 121 6.39 0.78 0.05
SP SW: C/C++ compilers   1 1 4 2 25 46 79 6.37 1.00 0.11
SP SW: Applications software     3 2 3 32 27 67 6.16 0.98 0.03
SP SW: General tools and utilities     4 3 8 37 34 86 6.09 1.02 0.18
SP SW: Performance and debugging tools 1   3 3 10 40 30 87 6.00 1.10 0.16
SP SW: Visualization software 1 1 2 4 2 18 12 41 5.67 1.47 0.27

 

Comments about Software:   15 responses

  • 6 overall software comments
  • 2 comments by Grid users
  • 4 comments by Jacquard Users
  • 2 comments by PDSF users
  • 7 comments by Seaborg users

 

General Software Comments

  6 responses

It would be nice to have ncdump and nco (netcdf operators) installed on the NERSC computers.

NERSC response: netcdf and nco are installed on all NERSC computers (Seaborg, Bassi, Jacquard and Davinci). To access the ncdump utility and/or nco (the netcdf operators) first load the module "netcdf".

I'd like to see LaTeX on a platform that can access the new global file system.

NERSC response: To request software for a NERSC machine please fill out the software request form.

Can I use Matlab? It's important for me.

NERSC response: MATLAB is available on Jacquard and Davinci.

I would like to have CLAPACK available

NERSC response: The LAPACK fortran library is available on all NERSC platforms in vendor provided libraries. LAPACK routines can be called from C-language programs; make sure that you follow the mixing Fortran and C rules on each platform. Note that CLAPACK (which uses a Fortran to C conversion utility called f2c) would have lower performance than LAPACK, which is why we recommend that C users use LAPACK.

... Although I have only used them slightly, some of the NERSC-developed performance analysis tools look to be helpful in increasing code efficiency

Our research focuses on demanding numerical problems in computational structural biology. In this regard, we would suggest two improvements:
It would be great to get the molecular dynamics software NAMD working better. We have successfully used it on seaborg, eventhough some inexplicable errors have occurred  On Jacquard, we have so far been unable to use it, despite communication with the consulting team.
Furthermore, adding the Poisson-Boltzmann PDE solver APBS to the list of available software would be of interest to us and possibly the larger molecular modeling community.

 

Comments by Grid Users:   2 responses

File transfer using grid resources (GLOBUS) is clumsy. Common tasks like recursively copying a directory structure don't seem to work. it seems fine for copying large single files, but not more complicated tasks.

The GRID-based parallel transfer of files is very useful but still awkward to use.

 

Comments by Jacquard Users:   4 responses

There have been major problems running the CAM climate code on Jacquard, that appear to be system related. Apparently there is a fix, but trying to test CAM on the development portion of Jacquard has been slow because there have been many issues related to a spartan environment (tools not installed, different versions of libraries, etc).

NERSC response: The kernel bug that affected CAM was fixed on Jacquard on Dec 13, 2005. You should now be able to run CAM on Jacquard.

I would prefer to use compilers from Portland Group.

A parallel debugger on jacquard would be of great help.

Need hardware performance monitors on Jacquard and Bassi.

 

Comments by PDSF Users:   2 responses

We need a group account so we would not be dependent upon individual users having to run group jobs. This is a major issue for us.

I am very happy with the recent decision to switch to GPFS. We already noticed much better performance and look forward to have the complete center switch to this file system.

 

Comments by Seaborg Users:   7 responses

It is appreciated if Matlab is available on seaborg.

NERSC response: MATLAB is available on the Linux Cluster, Jacquard, and the SGI Altix, DaVinci. We can't easily justify installing MATLAB on Seaborg because big MATLAB jobs would affect other users on the interactive nodes. DaVinci, which is tuned for interactive use, is the best platform for MATLAB.

Add ferret on seaborg

NERSC response: FERRET 5.81 is available on the SGI Altix, DaVinci. To access please First load the module "ferret". This version is not yet available for AIX, so it is not on Seaborg.

I have been very happy with the development environment at NERSC, particularly on Seaborg, although the relative leniency of its compilers in their default settings sometimes allows errors to persist that only get caught on other machines. I am glad to see that Bassi will be so similar.

Debugging is still a daunting task on Seaborg. I avoid doing it and, thus far, have not faced a situation where I have needed to do anything complicated (fingers crossed for the future!). GUI interfaces are frustrating for remote users; debugging with large numbers of processors is always awkward and frustrating within a batch environment. This is not a NERSC-specific issue. However, some innovation on this problem would be welcome. I rue the day when I find myself with a code that runs just fine on 1000 processors, but fails on 2000! ...

Since batch jobs had to wait too long in queue, I am wondering if in the future each use is limited to submit/run two jobs at one time so that every user has a chance to get job run at any time.

I'm not enough of an expert to comment on this intelligently. All compilers look alike to me (though I'm aware they don't to my codes).

It is frustrating to not have good syntax highlighting in VIM. The default terminal you provide isn't good, and the ansi terminal isn't much better. Don't you have enough people using linux to support a linux color terminal?
Can't you set up some sort of passwordless ssh so that it isn't such a pain to check many directories in and out of CVS?
Can't you tell your linker that it doesn't matter if symbols are unresolved if they aren't actually used? That is a gigantic pain and prevents some of our test code from compiling because some *UNUSED* symbols in a linked library aren't resolved.
Can't you fix your password checking? For a long password, where "a" is an alpha character, "s" is a symbol and "d" is a digit, aaaaaaaaaaasdd is not recognized properly, but sddaaaaaaaaaaa is recognized as satisfying your password security requirement policy.