NERSC 1998 User Survey Results

These are the complete responses to this question, with a few minor edits. Some responses have been repeated (to group them under multiple categories).

Provide more cycles / more machines:

• Upgrade J90 to handle vector programs when C90 is decommissioned.

• I would like to see the C90 kept. I guess it is expensive to maintain, but it is so much faster than the J90's, it's a real shame it must be mothballed.

• Continued availability of the C90 would be convenient.

• A new PVP machine with performance in the T-90 class or beyond to replace the C-90 when it gets phased out. The PVP machines are the lifeblood of the center.

• Have more moderately priced multiprocessor SGI, DEC, etc. computers available. Don't depend entirely on T3E type machine.

• Massively parallel vector machine. At least 1024 processors

• More MPP platforms

• More nodes on the T3E, another parallel machine.

• Need more parallel processing power.

• More massively parallel computers with MPI and good F90 compilers.

• The demand for computer cycles on massively parallel machine has far outstripped the supply.

• Our research depends very much on HPCC technology. More computing resources available to us will allow us to complete century long simulations and comparisons.

• Nothing important that I can think of! (Aside from many more node-hours --)

• more of the same!

• more computer power (obviously)

• Always more time.

• More of same.

• increase allocations

• More cycles

• Multithreaded Perl, but ESPECIALLY full ports of CERN libraries, as well as GNU tcsh (as opposed to clunky bash shell); full ssh to/from with authentication, to replace the disabled rlogin on T3Es; and Concurrent Versions System (CVS), an indispensable GNU tool for software version control.

• The NERSC computer system should acquire the MOLPRO program for high-level ab initio MO calculations. The MOLPRO program of P. Knowles and H.-J Werner has superior subroutines for advanced techniques such as the Restricted open-shell coupled cluster method which is more reliable and faster than that of the Gaussian version).

• Provide some tool for converting from FORTRAN 77 code to FORTRAN 90 code (like the VAST compiler) would be very useful.

• I would like to see the cf77 compiler remain available on the C90. My reason for this is that I have some codes which I have benchmarked on C90's with both the cf77 and f90 compilers, and I find that the f90 compiler is not very well optimized. In extreme cases, I have found the run time under the f90 compiler to be as much as 4 times slower than the cf77 compiler. Hence, I would like access to the cf77 compiler until the f90 compiler has received improved optimization.

• Sorry, but 100% compatible Fortran77 is a must for scientific applications. Not all our code is parallel and rewriting of some old but still useful pieces is a real trouble (like I/O procedures). Any sequential F77 would be ok for our group.

• install CERNLIB, this is a library of software packages for high energy physics research, but also useful for other works. [PVP user]

• Newer version of emacs on mcurie.

• ghostview utility

• file manager, editor, (network file synchronizer???)

• would you please provide the ezget libraries which Livermore people use to retrieve drs file?

• restore the pop mail server on SAS

• Additional batch queues would be desirable. Having to wait two days for a batch job to start then have it run for an hour or so is disappointing.

• Need longer time limits on T3E queues.

• I'd like to see NERSC provide a set of Unix scripts on line for making batch file submission easier and more user friendly.

• I am wondering whether the T3E scheduler could predict when a given job in the queue would be expected to start execution (this prediction could perhaps be included in the qstat listing); this prediction of course would be subject to change, but might be helpful for planning.

• more extensive web information would be nice

• More online documentation that "normal" people can read.

• good manual

• More training classes that are easier for people from remote sites with limited travel funds

• More information regarding libraries (math, graphics, etc) available and a friendly way to use such a libraries.

• additional space in the working disc [C90 user]

• more inodes for [T3E]

• Improved distributed use of facilities. For example, improve access to visualization server.

• machines with access to mcurie home directories with matlab

• When our group arrived in Berkeley, one resources that would been quite useful would have been some access to some DLT drives. We had many Gbytes of info on DLT tapes we wanted to bring in the J90 and HPSS, and the only network option would have been to transfer our data from Canada using FTP, and that would have taken days (if not weeks) of continuous transfer, since far away network connections like that are relatively slow. Instead we found some DLT tapes from a a group on the UCB campus (and transfer then using FTP or a short network distance) It would have been convenient if NERSC would have had some tape drive facilities for us to use locally. I'm not sure if the demand (probably very low) would justify the money investment for NERSC, but that's what resource I would have liked to have last summer.

• current updates about system, such as changes in the C90 compiler

• A more user friendly set of notes, e-mails, or video on the web regarding the current status of computers, availability of software, hardware, etc. This must be done in a way that reach the user, traditionally NERSC was the main source of computer power but this has changed and without another channel to update the users about what is going on at NERSC the tendency is for us to forget about NERSC and concentrate in other computer resources closer to us.

• A large linux farm.

• Farm of linux boxes

• PDSF should be supported strongly with more manpower and more machines.

• Our project generates large amounts of 4-dimensional data (3 space dimensions + time). We would like to have good, efficient means for analyzing and viewing this data remotely. Perhaps this capability exists and we don't know about it, or perhaps it doesn't exist. I'd like to know more.

• something like g-sharp -- together with avs .. visualization is great ... however, journal still want high quality black and white graphics [most of the time] and I have to publish my results in journals.

• Improved distributed use of facilities. For example, improve access to visualization server.

• Have experts in areas of user interest available as resources.

• No specific request at this moment yet.

• kaffee mit croissants

• No.

• No strong opinions.

• No opinion.

• I cannot think of any.

Back to Survey Summary Page