On Sat, Jan 3, 2015 at 8:53 PM, Anne Schilling <a...@math.ucdavis.edu> wrote: > Hi William, > > I have used combinat in the past for big research computations, but > not very recently (and I think I might have brought down the machine > at some point due to too much memory use). So as long as we can still > use the machine when we have large computations, I am ok with your plan. > > > Could you please be very specific what we need to do to get access > when running large computations? We would need to log onto SMC and then > e-mail you to move the process over?
That's precisely what I was thinking... except instead of "process" it's a "project", which is basically the same thing as a Linux account. For combinatorics researchers I would set the timeout of the project to infinite, and raise other quotas. Also, once you have a project you can directly ssh into "it" (using any ssh program) if you want, which in this could would literally just mean ssh'ing into an account running on a big virtual machine on combinat. One plus of this whole approach is that it's really easy to add your students/collaborators without my intervention. Once you have a "sage-combinat" project, if you want to have some students work on something there, just have them make SMC accounts then go to the project's settings and add them under collaborators (or just add them by email and they get an invite). I don't even have to be involved at all. A big problem probably with combinat.math.washington.edu, is that in practice people often want their grad students to do the real work :-), and it's a pain to get them accounts on combinat.math. In fact, I haven't made any accounts on combinat.math for anybody in months, and I'm not even sure currently if I can easily do so. Amusingly, in a typical day, the number of new accounts on SMC is almost the same as the total number of accounts I've ever given on that *.math.washington.edu compute cluster. Another SMC benefit is snapshots every few minutes, so if you accidentally delete a bunch of files, you can easily recover them... > I might still have some data/programs on the machine that I have not backed > up. I'll reboot combinat.math on Monday. Any files you have in /home/username are actually stored via NSF in a separate location, not on combinat.math, so they won't be impacted. William > Best, > > Anne > > On 1/3/15 8:23 PM, William Stein wrote: >> Hi, >> >> The computer combinat.math.washington.edu is down... again. Sort of. >> It responds to ping requests, but I can't ssh in. >> >> I suspect that not a lot of people are actively using it lately, since >> this is the second time it has gone down for over a week in the last 3 >> months, and nobody (except my student Hao Chen), seems to have >> noticed. >> >> I'm considering doing the following. I'll shutdown combinat >> completely, reformat the disk, and set it up as a node of the >> cloud.sagemath.com (SMC). It'll still have the amazing 64 cores and >> huge (192GB) RAM. However, instead of login in directly to it, people >> can email me to request that I move a particular SMC project to >> combinat. It will then have access to expanded compute resources. >> The advantage of this, is that it is much easier for me to maintain. >> In particular, SMC has automated scripts to take care of using cgroups >> to explicitly limit usage of compute resources by a given project, I >> have extensive monitoring code in place so I know when things go down, >> and I everything runs in virtual machines, so when there are problems >> I can easily fix them in a few minutes remotely. Also, it's much >> easier to grant fair usage to projects. As it is now with default >> linux on combinat, basically any user can just bring down the computer >> by using too much memory/disk/whatever, which is probably what >> happened in this case (I don't know). >> >> Thoughts? >> >> Obviously, this may be a bit slower and the max memory will be less >> (as things are in a VM) for specific research-level computations. >> However, a working computer is way better than a regularly-crashing >> computer, in my opinion. Also, given the weeks of downtime that >> nobody (except Hao) notices, maybe people aren't using combinat at all >> anyways, due to it being only a remote linux box. Personally, I >> think SMC makes using remote Linux boxes much easier. >> >> -- William > > -- > You received this message because you are subscribed to the Google Groups > "sage-combinat-devel" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to sage-combinat-devel+unsubscr...@googlegroups.com. > To post to this group, send email to sage-combinat-devel@googlegroups.com. > Visit this group at http://groups.google.com/group/sage-combinat-devel. > For more options, visit https://groups.google.com/d/optout. -- William Stein Professor of Mathematics University of Washington http://wstein.org -- You received this message because you are subscribed to the Google Groups "sage-combinat-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-combinat-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-combinat-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-combinat-devel. For more options, visit https://groups.google.com/d/optout.
