12.5 Moving data to a new computer

Lucky users who get a new computer frequently ask “how do I move my data to my new, better machine?” The easiest thing to do is first to install  on the new computer, release 31DEC07 or later. Create at least as many data areas on the new machine as there were on the old. If you are moving between machines of the same byte order (same architecture or Solaris to Mac PPC, Linux to Mac Intel, or vice versa), then do a network scp copy (or easier yet a cp from cross-mounted disks) to move all data files from  disk 1 on the old machine to  disk 1 on the new machine. Repeat for the other  disks. Note that  disk 1 is different; it contains message, SAVE/GET, and other special files as well as normal data files. The other data areas may be rearranged if you want, but do not copy two old data areas to one new data area. That will cause extreme confusion and loss of data. Now you are ready to resuming ing.

But if you are moving from Solaris or Mac PPC to Linux or Mac Intel, or vice versa, then you have more to do. Mount the  disks of the old machine from the new machine if possible. Otherwise copy the data from the old machine to new, temporary data areas on the new machine (not the new  data areas) being careful to keep each data area separate. Outside of AIPS, setup the  environment:

Answer the question about range of user number (0 0  C R will get you all possible numbers) and then enter the full pathname of the input (old) data area and the corresponding new  data area. The program will convert the format of all files. The program — barring software error — will produce correct output for almost all files. There is a small danger that the SAVE/GET and TPUT/TGET files may have some errors since it is way beyond the scope of REBYTE to understand the format of these files.

The traditional way to move between computer architectures is more time consuming. On the old computer, use procedure WRTDISK to write a data area out as FITS files. The FITS file names encode the old  disk and catalog numbers so that more than one disk can be written into the same FITS area. Copy the files to the new computer (or better yet, cross mount the disks) and then use procedure READISK on the new computer to read them into the new data area(s). These procedures are obtained via a RUN WRTPROCS. Plot and slice files are lost in this process. You can move the SAVE/GET files with the verb SG2RUN which converts an SG file to a text file which can be RUN on the new computer and then SAVEd. This traditional method works fairly well and can deal with the new computer having fewer  data areas than the old one. It also offers the opportunity to do a probably overdue backup of your data. However, the loss of message, plot, and slice files and the added overhead in disk space may matter to you.