there has been a large increase in the number of jaquith users lately.
here are some of the most frequently asked questions, in no particular
order.  possible future updates will also be sent to this aliase.  if
you do not wish to be included on this list, let me know.


1.  why is jls only listing the last bunch of files archived?

    jaquith has this concept of snapshots.  a snapshot is what
    a directory looks like at a particular point of time.  by 
    default, jls lists the most recent snapshot.  to see all 
    the files, use "jls -all".


2.  how do i get jls to list the last versions of all files, even
    across snapshots?

    try:        jls -l "*"
    double quotes are there to prevent shell from expanding star.

    there could still be bugs left in the indexing code.  but these 
    are minor inconveniences that shouldn't affect data integrity in 
    any way, so don't be scared :)


3.  where is my data going?  who's the server?  what do i have to
    do if i'm on s2k?

    your data is stored on the exabyte jukebox at berkeley.  the 
    default server is covet.cs.berkeley.edu.  if your machine is
    connected to berkeley via the s2k net, you should
	setenv JSERVER covet.s2k.Berkeley.EDU
    the server is running 24 hrs a day.


4.  what's exabyte's capacity?

    5GB per tape.  116 tapes.  4 readers.   
    currently, approximately 1/3 of the tapes have been claimed by
    various users.  

    tapes are assigned on demand, i.e. greedy users get the most.
    currently the ucsb folks are the clear winner.  the ucsb and esrg
    archives have chewed up a combined 11 tapes (unreplicated).
    sequoia researchers are encouraged to load as much data as quickly 
    as possible.   if it does fill up, we'll try to bring up something 
    else.   


5.  is there a file size limitation?

    yes, try to keep your files under 300MB.  if this is a problem, 
    i.e., you have lots of biggies, talk to me.


6.  where are the man pages?

    the man pages are installed on the sequoia machines.  if you
    don't have accounts on the sequoia machines and i didn't give
    you man pages, talk to me.


7.  how safe is my data?

    if you requested replication, i should've given you a "jput2"
    binary.  jput2 is not documented.  it has all the features of
    jput execept it replicates data in two archives.  use jput2
    in place of jput if you want some insurance.  and if you do,
    the chance of losing stuff is virtually nil.

    on rare occasions, tapes are munged due to overuse.  when that
    happens, you won't be able to jget.  if your data is not 
    replicated, then sorry :(  if your data IS replicated, then
    just let me know.  i'll bring up the secondary and back up the
    backup and you are ready to go again.

    the bottom line is your stuff is pretty safe.

    (older users that were given access to 2 archives and were
    told to jput twice can ignore jput2 if you choose to.)


8.  can jaquith files be shared between different machines, with
    or without shared home directories?

    yes.  jaquith only cares about names; it doesn't care about
    machines or inodes or whatever.  more specifically, let's 
    say you jput foo in /tmp on machine a.   then you jput bar
    in /tmp on machine b.  you can go to /tmp on machine c and 
    type jls -all and you should see both files foo and bar.


9.  what's jaquith's caching policy?

    files that are jput'ed start their lives on a 1+GB
    disk, therefore your jput's should be fairly fast.  when you
    jget files that are not on disk, they have to be fetched
    from the jukebox and such fetched chunks are managed are 
    managed using an lru policy.


10. my jget's are slow.  does it buy me anything if i start
    a bunch of them simultaneously?

    not really, your requests are queued on the same tape drive and
    serviced one after the other sequentially.


11. any suggestions to counter the slowness?

    prefetching wins if you can find useful things to do while
    waiting for your data.  here's how to prefetch with jaquith:

    + you start jget'ting file20-39 IN THE BACKGROUND;
    + do computation on file00-19 you have already fetched;
    + at the end of the computation, you wait until all the jget's
      have finished (by checking the file size of file20-39,
      for example);
    + repeat the process all over again.
    

that's all (for now) folks!
happy jput'ting,

Randy