I had run a job on workflow.isi.edu , and for the kickstart record the memory values seem inconsistent to me
bash-4.1$ grep rss impute2_ID0000017.out.000
<usage utime="12485.453" stime="19.805" maxrss="14320256" ixrss="0" idrss="0" isrss="0" minflt="3583101" majflt="3" nswap="0" inblock="1171568" outblock="6065312" msgsnd="0" msgrcv="0" nsignals="0" nvcsw="1300" nivcsw="19993"/>
<usage utime="0.010" stime="0.034" maxrss="828" ixrss="0" idrss="0" isrss="0" minflt="1447" majflt="1" nswap="0" inblock="24" outblock="0" msgsnd="0" msgrcv="0" nsignals="0" nvcsw="2" nivcsw="9"/>
<procs total="513" running="1" sleeping="511" zombie="1" vmsize="29927116800" rss="3580788736"/>

The value in the usage element maxrss 14320256 is in KB i.e around 14GB

However, the values mentioned in the procs element under the machine element appears insanely large if that parameter is in KB

<machine page-size="4096">
<stamp>2015-06-24T20:57:01.393-07:00</stamp>
<uname system="linux" nodename="workflow.isi.edu" release="2.6.32-504.16.2.el6.x86_64" machine="x86_64">#1 SMP Wed Apr 22 06:48:29 UTC 2015</uname>
<linux>
<ram total="67526049792" free="18273697792" shared="0" buffer="773070848"/>
<swap total="33839640576" free="33232605184"/>
<boot idle="127434677.100">2015-04-23T12:52:34.273-07:00</boot>
<cpu count="24" speed="2000" vendor="GenuineIntel">Intel(R) Xeon(R) CPU E5-2620 0 @ 2.00GHz</cpu>
<load min1="1.18" min5="1.68" min15="1.65"/>
<procs total="513" running="1" sleeping="511" zombie="1" vmsize="29927116800" rss="3580788736"/>
<task total="632" running="1" sleeping="630" zombie="1"/>
</linux>
</machine>

The vmsize="29927116800" rss="3580788736"/> if in KB would mean around 3TB for rss.. that does not add up.