Posts by dgnuff
|
41)
Message boards :
Number crunching :
Servers still down
(Message 59564)
Posted 15 Feb 2009 by  dgnuff
Post: The front page has the following note: February 16, 2009 A fileserver/SAN hiccup caused the entire project to pause for a number of hours. Sorry for the outage. -KEL However, as of now ( 9:20 AM 2/16/2009 PST ) the servers are not responding. A quick packet sniff shows that I'm sending packets with just the SYN bit set in the TCP header, and getting no reply at all. These are headed for 140.142.20.107, 140.142.20.111, and 140.142.20.112. Which implies the systems are down. Any idea of an ETA for return to service? |
|
42)
Message boards :
Number crunching :
Problems with web site
(Message 57414)
Posted 1 Dec 2008 by dgnuff
Post: Same here. I assume that the transfer of files will complete some time in the next day or two. The earliest deadline I have on a task awaiting upload is 12/6, or about five days hence as I type this. Personally, I'll keep crunching for Rosetta no matter what. However, I suspect there will be a huge uproar from other users who are more interested in statistics than I am if the delay becomes so great that tasks go past their deadline. |
|
43)
Message boards :
Number crunching :
Cheap dual-core build on NewEgg for R@H
(Message 56019)
Posted 25 Sep 2008 by dgnuff
Post: A couple of random thoughts on the topic. IME quad cores are far cheaper overall than dual cores. Mostly because you get to amortize a single mobo and case across more cores. I allocate 512 MB ram / core on my crunch boxes, so RAM comes out a wash. However, power consumption is the final killer for dual cores. I bought a kill-a-watt meter from ebay, it has been one of the best investments I ever made. In particular, it reports that my dualies consume 80w when running, while the quad cores only take 110w. If you do the math, that's 40w/core vs 28w/core, a clear win for the quads. That also means you do not want to pony up a bundle of cash for a case with a 400w PSU, you're just not going to use it. I use these cases from ebay. 200w PSU, and $25 delivered to my door. Since the kill-a-watt says I'm only taking 110w, 200w is plenty enough. If you're going the Linux route, take a look at how to use pxe to boot them over the network - just about all modern mobos will support a network boot. |
|
44)
Message boards :
Cafe Rosetta :
Tell Wired magazine your story!
(Message 55861)
Posted 18 Sep 2008 by dgnuff
Post: According to the WHO, 1979 was a landmark year, in that smallpox was eradicated in that year. That was a major accomplishment, because smallpox was a terrible disease, the Variola major variant causing death in about one third of cases. Compare that to HIV, which is having a devastating effect in parts of the world. Dr. Baker has a clear roadmap for work towards an HIV vaccine, crunching for Rosetta@home means I can be a part of that effort, and it gives me a shot, albeit a small one, of having the computer that actually finds the cure. |
|
45)
Message boards :
Cafe Rosetta :
Got a couple of systems for sale
(Message 55375)
Posted 29 Aug 2008 by dgnuff
Post: I'm slowly upgrading my crunch farm, which means that I need to let go of a couple of the systems in it. They are: 878465 and 878455. I'd like to get $75 each of them, plus shipping, or you can pick them up if you're in the SF Bay Area (Hayward). I'll wipe them and reinstall the OS, they do come with a legit copy of XP: they're OEM HP systems, I have the reinstall disk. I can also pre-install boinc on them if you want, as well as VNC (my choice for my headless systems). Drop me a line here if you're interested ..... |
|
46)
Message boards :
Cafe Rosetta :
my futile endevour to recruit
(Message 55278)
Posted 25 Aug 2008 by dgnuff
Post:
I bought one of these to measure how much power the systems in my "crunch farm" are using. The two dual core boxes (E2200s) draw 80 watts each, while the two Celeron systems draw 110 watts. These are scheduled to be replaced, I'm planning on building a couple of quad core boxes. 80 watts means a unit of electricity (one KWh) every 12 hours, or two units a day. Since I pay 13 cents / KWh, this works out to about a quarter a day for each of the dualies, and maybe 35 cents / day for each of the Celerons. Put this another way. If I turned off all four crunch systems, it wouldn't even pay for a daily latte from Starbucks. |
|
47)
Message boards :
Number crunching :
Which file holds the run time parameter ?
(Message 55277)
Posted 25 Aug 2008 by dgnuff
Post: *** Edit *** Deleted *** Read elsewhere that the problem will cease to be an issue in a month or so. |
|
48)
Message boards :
Rosetta@home Science :
Can we produce an amylase equivalent for cellulose?
(Message 54883)
Posted 3 Aug 2008 by dgnuff
Post: A search for "cellulose" produced nothing, so ... Ruminant animals have bacteria in the first two stomachs that are capable of breaking down cellulose into sugars. I can't help but wonder whether Rosetta is capable of designing a custom enzyme that could do the same thing in vitreo. The uses of such an enzyme as a precursor to fermentation for fuel alcohol would seem to me to be an interesting and possibly worthwhile research project ... |
|
49)
Message boards :
Number crunching :
How can I get the most useful information to Bakerlab in the event of a crash
(Message 54881)
Posted 3 Aug 2008 by dgnuff
Post: In the last month or two, I've had several cases where a WU crashes. All I get from Windows is a notification, and a dialog that lets me send information to Microsoft. I realize that's probably a complete waste of time, since MS most likely won't forward the information to Bakerlab. That said, when this happens is there anything at all that I can send directly to akerlab that will help them track down the problem? Thanks for any suggestions ... |
|
50)
Message boards :
Number crunching :
Team Points STOLEN!!!!! Over 6 Million Points!
(Message 54475)
Posted 13 Jul 2008 by dgnuff
Post: Just a couple of thoughts.
The part I've italicized is an out and out lie, since as Scribe has noted, there were still active Teddies crunchers here at the time of the theft. Secondly, it seems to me that the obvious place to fix this is home base: http://boinc.berkeley.edu Are the main Boinc development team aware that this is going on? An almost identical problem exists in online games, an area where I'm very active. You'll get guilds in these where leaders go inactive. In almost all cases, leadership will automatically transition to another member who is active. To avoid a coup of this sort in the future, the approach I'd suggest is as follows. If the leader has not submitted results for more than <insert time here>, find all members that have returned results within that time period. Then promote the member from that set with the earliest join date. By giving preference to long time members, that should do a lot to prevent someone joining and immediately taking over. |
|
51)
Message boards :
Rosetta@home Science :
Rosetta vs POEM
(Message 54054)
Posted 28 Jun 2008 by dgnuff
Post: Jumping in here a bit late, but the following paragraph from POEM's page has me a little puzzled.
(Emphasis mine) The part that I've highlighted is what has me confused, because my limited understanding of what we're doing here appears to rely on precisely this property. |
|
52)
Message boards :
Number crunching :
OMFG!!!!!!! folding just broke the teraflop boundrie!!!!!
(Message 46751)
Posted 22 Sep 2007 by dgnuff
Post: As noted, PS3GRID have a Boinc client, so all that would be needed is to port the Rosetta client to cell linux for PS3 and we'd be a go. Yeah, I know. That's *INCREDIBLY* easy to say. Doing that in such a way that it uses the SPE's effectively will not be easy, largely due to the limited memory capacity of the SPE's. To the best of my knowledge, the basic idea used by the PS3 F@H team was to simply subdivide the work and brute force it. I believe that on a PC the main "atom to atom" computation is only done if a first pass determines that it's worth while. On the PS3, they took a rather more brutal approach: just try everything, and discard insignificant results after the fact. Even doing this, you can see the speedup they got. Which raises the question for the Rosetta team ..... Does the Rosetta algorithm work on the basis of attempting to compute a force/distance/whatever between every pair of atoms for a given comformation, or at least every pair of atoms where it makes sense? In the more general case, at a conceptual level at least, what exactly does the "inner loop" do? While I'm at it, is there a "cliff notes" overview of the algorithm, as in something that outlines its operation in five minutes or less of reading? |
|
53)
Message boards :
Cafe Rosetta :
Word link 10
(Message 46746)
Posted 21 Sep 2007 by dgnuff
Post: Iron men ... Wooden Ships |
|
54)
Message boards :
Rosetta@home Science :
Boinc/Rosetta on the Xbox 360?
(Message 43063)
Posted 2 Jul 2007 by dgnuff
Post:
I need to touch this one. You're 100% right, but for 100% the wrong reasons. The PS3 (cell) is far harder to code for, but it's because of the limitations of the SPU's, most notably memory. Each SPU has only 256K of local store, but it runs at effectively L1 cache speeds. So to get performance for Rosetta, we would need to take a WU, chop it into bite sized chunks which can run for a couple of ms on an SPU, while only using 80 to 100k of data. That allows double buffering of data: DMA out N-1 results, DMA in N+1 data, all while N runs. At 100k per WU, that's 200K gone, which leaves 56K for code and "incidentals". |
|
55)
Message boards :
Number crunching :
PS3 preliminary crunching numbers
(Message 43056)
Posted 2 Jul 2007 by dgnuff
Post: ** Deleted ** |
|
56)
Message boards :
Rosetta@home Science :
"How close are we?"
(Message 42120)
Posted 12 Jun 2007 by dgnuff
Post: Both the questions and answers in this thread are excellent. The reality is that there are a number of steps to go from a designed protein to an actual therapeutic or vaccine used to treat people. The most time consuming and expensive step are clinical trials to see if the designed protein works as a desired as a drug or vaccine and to verify that it does not have any undesired effects. We are vey excited about the potential of the proteins we are designing to cure diseases, and believe that the approach we are collectively taking to developing cures are among the best possible, but it will take time to first, design a vaccine or therapeutic protein that has exactly the right set of biochemical properties, and second, to prove that the new protein acts as desired without side effects in people. So a good analogy might be to think of this like producing a diamond for a ring. What we're doing is the initial search through the ground for those rare uncut diamonds. By harnessing the power of a large collection of PC's it becomes possible to reduce this search to something manageable. However even when we find something, a lot of work is still needed: this compares to cutting and polishing the gem before the ring is finally made. And not every uncut diamond can make a ring. Due to flaws in it, it may have to be cut into bagatelles that are used as highlights for something else. |
|
57)
Message boards :
Number crunching :
PS3 preliminary crunching numbers
(Message 42119)
Posted 12 Jun 2007 by dgnuff
Post: WOW! These numbers are crazy! The thing to realise about that is that the Sony FAH team wrote some very carefully crafted inner loops that run on the SPUs, which is where the PS3 gets those obscenely high flops counts from. That, of course, raises the question of whether the Rosetta algorithm is amenable to being parted out on a single system to the 6 available SPUs, especially considering that they each only have 256K of memory. Looking at the typical WSS under windows (30 to 50 MB, rising to 100+ MB sometimes), I'm not sure what the answer is. |
|
58)
Message boards :
Number crunching :
Is this a Boinc problem or Rosetta problem?
(Message 30029)
Posted 26 Oct 2006 by dgnuff
Post: I'm not sure which of the two it is, I tend to suspect Boinc.exe, but can't prove it. All I find is the following in stderrae.txt: ***UNHANDLED EXCEPTION*** Reason: Access Violation (0xc0000005) at address 0x0032f114 read attempt at address 0x00000008 Now, if this is a BOINC error, it should be noted that I'm still running 5.11, not having come across any overpowering reason to upgrade. This is only happening on one Windows machine, I've got three others all of which run without problems. The one in question is a P4 (Dell Dimension 4700), the other three are two "beige box" AMD 3500+'s and a Pentium M notebook. Any ideas? |
|
59)
Message boards :
Rosetta@home Science :
Food for thought - Flu vaccine
(Message 29822)
Posted 22 Oct 2006 by dgnuff
Post: After a friend got quite ill with the flu *FROM* taking a flu shot this year, I got me to thinking about it. As far as I know, flu is a virus (like HIV). I also believe that flu vaccine is currently created by taking the best estimate as to what the "strain of the year" will be, cultivating it, weakening it somehow, and then throwing it at us. Assuming the two statements in the previous paragraph are correct (I'm willing to be corrected if they are not), my thinking was the following. Does flu share with HIV the property that certain parts of it are immutable, while others mutate, thus generating the various strains we get each year? If so, can we attack it in the same way as the intended solution for HIV, i.e. to generate a custom protein that triggers the immune system to generate antibodies for the immutable parts of the flu virus. Doing so would have two advantages, as I see it. Firstly, we could use the same flu vaccine for all shots, thus removing the need to generate "per strain" vacines as we do now. Also, since it would not be a weakened form of the virus, it would be a lot safer for folks like my friend, or for me. I'm diabetic, so I have to live with a suppressed immune system. Think what this means in terms of even a weakened flu virus. Just some food for thought ..... |
|
60)
Message boards :
Cafe Rosetta :
Word Link 4
(Message 29671)
Posted 20 Oct 2006 by dgnuff
Post: Heart |
Previous 20 · Next 20
©2024 University of Washington
https://www.bakerlab.org