 |
| Stephen Brooks 2013-08-03 03:10:34 | I've accepted a job at BNL, home of the RHIC accelerator. I'll still be using Muon1 in the work for them, but on different future projects than the neutrino factory (an electron-ion collider or maybe an accelerator to drive an X-ray laser, or possibly a high-power accelerator to transmute radioactive waste, or actually there's also interest in a cancer-zapping accelerator for "hadron therapy" ). In fact BNL also have a neutrino factory/muon collider group, but that project seems further off in the future than certain others right now. What this means practically is that there'll be probably a couple of weeks server outage at some point because my computer needs to be shipped across the Atlantic. Before that there might be a few days when I have to reinstall the Muon1 server on my home computer from the work one. As for what work the network will do, there's still demand for optimisation of accelerators on these new projects and I've already adapted the code and am running some first experimental Muon1 optimisations on those totally new accelerators locally on my workstation. But it will be some months before I've moved and am in a position to release a new version plus the optimisations to go with it, but the staff here I've talked to are generally keen that I continue to make use of this project. |
| yoyo 2013-08-05 06:41:15 | Should we continue on the current lattices or are these result no more used? Should we stop until the new app and new lattices are out? yoyo |
| Stephen Brooks 2013-08-08 21:52:38 | You should continue to use Muon1: it may even be that the neutrino factory design work carries on in parallel to the new projects in the future. The only time I'd suggest stopping (and going to other projects temporarily) is when I'll have to switch off the stats server to do my move in ~October. |
| Silverthorne 2013-08-10 15:44:51 | Congratulations on the new job!! [Edited by Silverthorne at 2013-08-10 15:45:02] |
| K`Tetch 2013-08-10 20:39:04 | I'm sure we can fix something up to work temporarily, Stephen, for the stats server. Even if it's just me setting up an old computer for you here, and it running in a reduced mode. |
| Stephen Brooks 2013-09-18 18:58:33 | My final day at my current employer (where the Muon1 stats server is) will be October 4th. Currently the stats are stored on a 320GB hard drive, which is actually mine, although the workstation belongs to the laboratory. Currently 153GB out of 298GB logical space on that disk is used for the results database, with NTFS compression turned on. It is also backed up nightly onto a network-attached drive. I was going to just take this 320GB disk out and install in my personal desktop computer along with the stats software and run from home (and later, the US) but I decided now was a good time to buy some extra storage: 2TB Seagate ST2000DM001 Barracuda 7200.14 SATA 3 6GB/s 7200rpm 64MB Cache 8ms OEM NCQ I've ordered the above disk and will copy the database over as I'm doing the transition from work to home. |
| RGtx 2013-09-19 00:54:17 | 2 months ago, my new 2TB Western Digital WD2002FAEX Caviar Black died within 24 hours of installation, so don't be too hasty in wiping your old HDD. |
| Stephen Brooks 2013-09-19 16:40:24 | No, the plan was to actually leave the 300GB in the system as a nightly mirror, at least until the dataset gets too big for it. I'm already doing that with my main 1TB drive at home onto an old 300GB drive (fortunately my PC has 4 slots for hard disks). [edit] Today I've been moving various scripts onto the stats hard disk so all the programs required for running the project are in the same place. That meant some recompilations and changing the paths from absolute to relative in the programs. What this means is: watch out for oddities in the samplefiles, stats or graphs. If they don't refresh properly for an extended period or display junk data, let me know. [update] Most of them work in their new home. The samplefile generator had printed the error message "LONG BENT DUCKS" and was pausing for 12.4 days, which means I hadn't set the current working directory properly. Hence no samplefiles yesterday. They should update soon. |
| Stephen Brooks 2013-09-20 23:56:41 | Just a message to say I'm advancing the schedule for moving the stats server to my home computer: I'll try and do it on early next week (23rd-26th). It will be down most of the day and possibly the next, depending how well it goes. [edit] A potential problem is that BNL seems to block FTP, so I'll have difficulty running the project while I'm housed on-site. I'm asking if I can get an exception. Otherwise we may try to convert the servers to SFTP?? Next week I need to test if I can get uploads to my website via WebDAV, which is their only non-FTP method. But getting from the FTP servers to the stats server is still tricky. That should cease to be a problem once I've moved out into normal accommodation but that will probably be in December. |
| Stephen Brooks 2013-09-23 18:00:03 | Enquiries to the lab say I need to use an "FTP gateway". Not sure what that involves or if I need to change my scripts. |
| [TA]Assimilator1 2013-11-09 01:04:54 | Ah that's why you're moving, congrats  Btw what is with you & ducks?  lol..... what was that smilie? :duckrot - nope [Edited by [TA]Assimilator1 at 2013-11-09 01:05:08] [Edited by [TA]Assimilator1 at 2013-11-09 01:05:43] |
| [OcUK]diogenese 2013-11-09 05:29:17 | Do you mean [img]http://stephenbrooks.org/forum/smilies/duckrot.gif[/img] ? That didn't work, it's something like that anyway!  [Edited by [OcUK]diogenese at 2013-11-09 05:30:43] |
| Stephen Brooks 2013-11-09 18:04:15 | It's  |
| Stephen Brooks 2013-11-09 18:06:17 | No, looks like it's not. I think it's (°)> |
| Stephen Brooks 2013-11-09 18:06:53 |  |
| Stephen Brooks 2013-11-09 18:07:14 | (°> |
| Stephen Brooks 2013-11-09 18:08:38 | What is going on with this duck (°)> |
| Stephen Brooks 2013-11-09 18:14:14 | Well, I've given up trying to get PHP to accept the extended character ° in (°)> in its search and replace functions. Use or in future. |
| Stephen Brooks 2013-11-09 18:15:45 | Oh hang on, it was a problem with the beak. The beak is an HTML entity. That should fix it for duck classic: (°)> |
| Stephen Brooks 2013-11-09 18:17:07 | No, seems it still hates extended characters. Back to and . |
| [OcUK]diogenese 2013-12-10 08:20:16 | What is the timeframe for new work relating to the brookhaven accelerator? Sorry about the extra posts, I don't know how to delete them [Edited by [OcUK]diogenese at 2013-12-10 08:23:43] |
| Stephen Brooks 2013-12-10 21:04:34 | Well I've got a few optimisations running here on it already. Not sure how stable the client is with the new features. Also some of these produce a large number of small results e.g. 100000 per day with 0.1 Mpts each, so it might kill the stats server to have that much data coming in. They finish quickly because I'm using a mode with only about 10 test particles. There is eventually going to be a need for simulating with a full distribution (10-100k like in the neutrino factory simulations) and for that I might put it on the network. |
| RGtx 2013-12-11 01:53:47 | That's an awful lot of HDD activity, or are you running it on a virtual RAM drive? |
| Stephen Brooks 2013-12-11 15:02:21 | It's fine for a single client (2-3 simulations per second) but it would kill the stats server. I either find a multiparticle application that's useful, or another way is put "results merging" in the client, so it combines say 1000 results into the best 5 while adding up the Mpts. There's also some compensation to be done in Mpts for that case because the initialisation takes a large chunk of the time (rather than the simulation) and at present that isn't scored. |
| Stephen Brooks 2013-12-27 20:54:15 | I don't think I've mentioned this in other threads but last week I found the switch responsible for solving the slow MinGW gcc problem. -O3 does a little bit but -ffast-math does a lot because it abandons strict adherence to IEEE floating point, saving time. The machine architecture flags (SSE2, Pentium 4 etc.) don't seem to do a lot for my programs. Though this was still gcc 3.4.2 and 4.8.1 is available as a MinGW port too, which I've not tested yet. |
