<html>
<head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<br>
Food for thought.<br>
<br>
I've been following this thread with some interest because<br>
I had been wondering how best to handle SGI's machines<br>
in a remote beaker lab.<br>
<br>
There are a few fundamental problems:<br>
1) The machines are sometimes unique -- only one exists, period.<br>
The result is that at certain times the machines must be
available<br>
to particular users -- customer/trade show demos for example.<br>
2) They often require hours of continuous time which, particularly<br>
during bring up and major OS releases means scheduling<br>
evenings and weekends --- engineers are arranging their sleep<br>
schedules to take advantage of certain time slots. Random<br>
start times for reservations won't work.<br>
<br>
What seems to be lacking in beaker is <b>Hard Start Times</b> which<br>
implies hard-end-times.<br>
<br>
So for example, if we put a 4 rack UV system (2048 cores, 16Tb of<br>
memory) in beaker it is not just going to stay there; it will be
pulled<br>
back inside SGI's firewall on a regular basis.<br>
<br>
I was hoping that reservations would address this issue.<br>
<br>
Thoughts?<br>
George<br>
<br>
</body>
</html>