Posts by ivan


log in
1) Message boards : Number crunching : How To Do A "Headless Linux CLI Multiple GPU Boinc Server" Semi-Pro Cruncher ;) (Message 1613238)
Posted 7 days ago by Profile ivan
I'm also interested.

Faster graphics cards would of course be better. I know nothing about rack cases. Do they make them taller so you could put full height cards in them?

Yeah, there are a number of different configurations. One of my Supermicros is only 1U high but it has a C1060 and a GTX460 (sans fan!) in it -- they lie flat so height isn't really an issue. Similarly with my latest 2U chassis -- it has two slide-in dual-Xeon systems, each of which can take a full-size PCI-e card; we have the same chasses in our Grid cluster with four smaller slide-in dual-Xeons, but they can only take low-profile cards (like the GT640 with appropriate hardware). One of the other groups in our Department tested an Intel system a while ago that could take eight (IIRC) PCI-e cards, standing up, but I don't remember the chassis size or whether they were full-height or low-profile slots.
2) Message boards : News : Database Rebuild Status update. (Message 1611451)
Posted 10 days ago by Profile ivan

WEATHER CANBERRA Daily forecast
Day Weather Maximum Day Temperature Minimum Night Temperature Wind Direction and Speed
Tuesday Sunny Interval 25°C 16°C East North Easterly 4mph
Wednesday Light Rain 27°C 18°C North Easterly 3mph
Sunrise 05:42 Sunset 20:10


Curious Ivan you have a pommy flag and you Quote the weather in Canberra ?

Working in our nations capital mate ?

Don't worry about the U.V summer has only just started mate , just wait till January !!

Just visiting, Australian Institute of Physics Congress (gwgl AIP2014). Then back home to Byron Bay for the rest of the month before returning to London (any local s@h-ers in BB feel welcome to contact me).
3) Message boards : News : Database Rebuild Status update. (Message 1610934)
Posted 11 days ago by Profile ivan
Ugh, these people have the luxury to only work when the sun is up? Bah!

When I'm at work at 8am this time of year at the Northern Hemisphere, there's not been a sun in my life yet! If it wasn't dreary as today, it usually is down as well when I go back home at 4.45pm. I'll have to wait till end-of-January to be able to see the sun still at the end of the day.

;-)


Sunrise Tuesday 9th : 9:45am, Sundown Tuesday 9th: 2:55pm. Length of the day: 5 h 10 min.

And I wonder why people think that us Finns are grumpy people ;)


Kingman, AZ
Sunrise: 07:31:57 Sunset: 17:22:52 Transit: 12:27:24 Length of daytime: 9h 50m 55s

WEATHER CANBERRA Daily forecast Day Weather Maximum Day Temperature Minimum Night Temperature Wind Direction and Speed Tuesday Sunny Interval 25°C 16°C East North Easterly 4mph Wednesday Light Rain 27°C 18°C North Easterly 3mph Sunrise 05:42 Sunset 20:10

So far I've not got too much UV...
4) Message boards : Number crunching : Panic Mode On (92) Server Problems? (Message 1606856)
Posted 21 days ago by Profile ivan
Down to one in-process job. Now have Einstein running on several of my "best" machines in the interim. I did have a problem with it on my work Win7 desktop -- a quad-core i7 (with hyper-threading) with 4 GB of RAM. It spawned eight tasks of 400 MB each, and sent the machine into swap territory! I eventually managed to regain enough control to be able to set local preferences to "use at most 50% memory" but it took a while! I'm away, imminently, for the next five weeks so I'm not overly concerned about interactive response for the nonce.
5) Message boards : News : Running out of workunits (Message 1606854)
Posted 21 days ago by Profile ivan
I'm going to blame the NSA, China and Russian's for the problems
November 5th I had a total system crash, damaged Windows and lost Linux Mint 17 including all the work units I had downloaded.
Everything had been running fine for months
I had to fit hard drive with a Linux Ubuntu 12.04LTS to get system running again. No chance of repairing Mint installation, may be able to repair Windows but OS now says I only have a single core processor instead of 4 core, BIOS still sees 4 core so that doesn't look damaged

Just a word of advice (ignore it if you already know...). On Linux, make the /home directory a separate partition. Then, if you have to change/upgrade the OS, your personal files remain intact (you have to do a manual disk allocation in the OS set-up of course). I used to also have a separate partition for /boot but when Ubuntu decided not to delete older versions of the kernel I got into awful trouble having to resize it manually (very manually) to get things working again. These days I usually have just three partitions -- /, /home, and swap. If it's a data-handling machine than a /data would also make sense. OTOH, there's the 11th commandment, Thou shalt make backups!
6) Message boards : Technical News : I was right... (Nov 24, 2014) (Message 1606321)
Posted 22 days ago by Profile ivan
Tuht mir leid wenn ich auf Deutsch Poste aber Mein English Reicht gerade für kleine Sätze aus und nicht für große also Translat this with Google Translator or Bing Thanks and Habby gobble Gobble Day from Germany now in German Wan wirds entlich fertig mit dem Umbau ?


Yep, he's right. This must be the worst ever Google translation from German to English:

Tuht'm sorry but if I My English to German wirds Poste enough, especially for small sets and not so great for Translat this with Google Translator or Bing Thanks and Gobble gobble Habby Day from Germany now in German Wan entlich finished remodeling?

Sheesh! Even I can do better, and I only lived in der Schweiz for 13 years:
"Sorry that I post in German but my English is only good enough for short sentences and not long ones -- so translate this with Google or Bing. Thanks, and Happy Thanksgiving from Deutschland. Jetz auf Deutsch; When will the rebuild finally be finished?"
Tja, niemand weiss. Wann es fertig ist!
7) Message boards : Technical News : What's happening... (Nov 8, 2014) (Message 1604124)
Posted 27 days ago by Profile ivan
Life is too short to drink cheap beer..........

Life is also too short to drink 11% wine! (California vintners, please note...)
8) Message boards : Number crunching : Panic Mode On (92) Server Problems? (Message 1603794)
Posted 28 days ago by Profile ivan
From the home page,

The science database server crashed a few hours ago. The project is down until the database recovery is complete.


Cheers.

Who woulda thunk to look there?

When I was a little kid, the general store in Byron Bay (back before self-service, you asked at the counter for what you wanted, and they fired your money and the bill on overhead wires up to the cashier above, who returned the little trolley with your change and receipt...) there was a sign on the counter, "When all else fails, raise this flap!" Should you actually raise the flap, you saw the message, "Not now, stupid, when all else fails!"
Given the reluctance of physicists to read the documentation, I've often thought of prefacing an instruction manual with, "When all else fails, read the instructions." With, of course, the obvious riposte on the first page...
[Edit] Oops, put that reply to the wrong post! [/Edit]
9) Message boards : Number crunching : Best performing hardware (Message 1602713)
Posted 19 Nov 2014 by Profile ivan
Then there's the memory problem -- our particular model only has 8 GB of ram, which has to contain the OS as well as applications. At the moment top reports 7.6 GB free, so for 60 cores x 4 threads that'd be only around 30 MB available per thread; currently on an Ubuntu box s@h reports 104 or 164 MB virtual memory per process, 40 or 96 MB resident per process, and 12 MB KB shared. (Similar figures on a RHEL box; the one job on Ubuntu that's taking more RAM is not the vlar WU running at the moment.)

If you are running stock apps, those are UPX compressed. Decompress them and see what figures you get then. Although, iirc, the apps use more than 30 MB for data.

Those are in-memory figures from top -- compression of the executable wouldn't affect that (only the space on disk), surely?
10) Message boards : Number crunching : Best performing hardware (Message 1602580)
Posted 19 Nov 2014 by Profile ivan
[quote]
Then there's the memory problem -- our particular model only has 8 GB of ram, which has to contain the OS as well as applications. At the moment top reports 7.6 GB free, so for 60 cores x 4 threads that'd be only around 30 MB available per thread; currently on an Ubuntu box s@h reports 104 or 164 MB virtual memory per process, 40 or 96 MB resident per process, and 12 MB KB shared. (Similar figures on a RHEL box; the one job on Ubuntu that's taking more RAM is not the vlar WU running at the moment.)
This might all change with the new Phis, which slot into motherboard sockets and can access main memory, but Intel hasn't offered me one to play with yet...

Would it be possible to use massively-parallel system via OpenCL drivers and not to run separate instance on each node/core ?

Perhaps, you have more experience there than I. I've done a few OpenMP things but not really had the chance to play with it otherwise. Do you have to explicitly parallelise everything in OpenCL or does the system work that out for you? (Guess I should dig out your code if I get an idle moment.)
Another possibility would be to try to get the shareable section much larger by using shared libraries:
[eesridr:~] > ldd BOINC/projects/setiathome.berkeley.edu/setiathome_7.01_x86_64-pc-linux-gnu not a dynamic executable
11) Message boards : Number crunching : HDD Questions For The Elite - [RESOLVED] (Message 1602303)
Posted 18 Nov 2014 by Profile ivan
[I don't do media sharing, at least I don't think I am. ;) Nothing is hogging a lot of resources. I'm still leaning toward the controller getting wonky. The light is not showing any disk access. It is "burning bright" just as my power switch is, no winking, flickering or anything, just solid on.

Keep on BOINCing...! :)

It could just be the driver transistor/gate gone short-circuit and pulling the LED permanently on. I had something similar with one of my new Xeons lately, the front panel indicators showed that the RAID array was continually being rebuilt. After replacing the disks, and the motherboard, it finally went away when we wiggled the SAS disk cables, so we replaced them too. Should have traced it down from the schematics, but they don't seem to provide them these days -- it's just "keep replacing components until the problem goes away."
12) Message boards : Number crunching : Best performing hardware (Message 1602280)
Posted 18 Nov 2014 by Profile ivan
Correct me if I'm wrong, but AFAIK the important thing for SETI is single precision performance.

The Xeon Phi 7120P seems to have a theoretical single precision peak of 2.4 TFLOPS/s:
http://www.intel.com/content/www/us/en/benchmarks/server/xeon-phi/xeon-phi-theoretical-maximums.html

A GTX780 or a Titan seems to be much faster:

The GTX 780 still offers respectable single precision performance though, clocking in at 4 Teraflops compared to the Titan's 4.5 Teraflops.

http://www.maximumpc.com/article/news/geforce_gtx_780_benchmarks

The Titan Black is rated at 5.1 TFLOPS/s
http://www.bit-tech.net/news/hardware/2014/02/18/nvidia-gtx-titan-black-launched/1

The GTX980 should be about the same with ~5 TFLOPS/s
http://www.pcworld.com/article/2686115/nvidia-unveils-its-all-new-geforce-gtx-980-and-gtx-970-graphics-processors.html

I have a Xeon Phi. I gave up on trying to port BOINC and S@H to it. For the best performance you have to run native code on the Phi cluster (otherwise communication bottlenecks between the host and the cluster slow you down). That would mean running BOINC on the Phi as well and BOINC code is not the most portable (but not too bad if you ignore boincmgr and use boinccmd for all interactions).
Then there's the memory problem -- our particular model only has 8 GB of ram, which has to contain the OS as well as applications. At the moment top reports 7.6 GB free, so for 60 cores x 4 threads that'd be only around 30 MB available per thread; currently on an Ubuntu box s@h reports 104 or 164 MB virtual memory per process, 40 or 96 MB resident per process, and 12 MB KB shared. (Similar figures on a RHEL box; the one job on Ubuntu that's taking more RAM is not the vlar WU running at the moment.)
This might all change with the new Phis, which slot into motherboard sockets and can access main memory, but Intel hasn't offered me one to play with yet...
13) Message boards : Number crunching : How to limit size of job_log_setiathome.berkeley.edu.txt (Message 1601179)
Posted 14 Nov 2014 by Profile ivan
Mine is 41 MB and 385000 lines

Over here they call this "willie waving"... :-)

[eesridr:BOINC] > wc job_log_setiathome.berkeley.edu.txt 469562 5165182 51731379 job_log_setiathome.berkeley.edu.txt [eesridr:BOINC] > head job_log_setiathome.berkeley.edu.txt 1273852033 ue 167272.344521 ct 7.204450 fe 167272344521056 nm 10fe07ah.31298.5798.12.10.173_0 et 9.966059 1273852035 ue 167272.344521 ct 6.804424 fe 167272344521056 nm 10fe07ah.31298.5798.12.10.191_0 et 8.589399 1273855954 ue 167272.344521 ct 3890.751156 fe 167272344521056 nm 10fe07ah.31298.5798.12.10.100_0 et 3934.674194 Convert Unix timestamp to Readable Date/time (based on seconds since standard epoch of 1/1/1970) UNIX TimeStamp: 1273852033 Fri, 14 May 2010 15:47:13 GMT
14) Message boards : Number crunching : Lunatics Windows Installer v0.43a Maintenance Release (Message 1601095)
Posted 14 Nov 2014 by Profile ivan
Is there a way to add all 3 cuda apps, and can this be done by editing the app_info script since you only have an option to check one of the boxes?

As far as I understand it, you don't want to do that. The stock app sends the different versions because it doesn't know what hardware you have; once enough units have been crunched it selects the most efficient and only targets that version (the work-units are the same, it's just the targeted version that can vary). By selecting the most appropriate version during Lunatics install you can short-circuit this lengthy and inefficient selection process.
15) Message boards : Number crunching : BOINC 7.4.27? (Message 1600164)
Posted 12 Nov 2014 by Profile ivan
http://boinc.berkeley.edu/download.php

???
Download BOINC for Linux x64 (3.25 MB) BOINC 7.2.42
16) Message boards : Number crunching : tripple gtx 980 water cooled setup (Message 1599822)
Posted 11 Nov 2014 by Profile ivan
Hopefully that will solve their chronic fan-quality problems! :-b
17) Message boards : Number crunching : hyperthreading and processor affinity in linux with taskset (Message 1596056)
Posted 2 Nov 2014 by Profile ivan
So I moved it to local storage ( SSD ). Then looked a little closer at it while it ran. Regularly ( every few seconds ) hitting 100-200 MB/sec of disk IO ( over 1/2 second averages as viewed via iostat/iotop). So this project was IO bound. Kept my network and server overtasked.

I've installed lm-sensors and gkrellm on as many of my Linux boxes as I can. Then run sensors-detect to collect sensor info and run gkrellm, configuring which sensors and subsystems to monitor. I have a script that starts up
ssh -Yf user@machine gkrellm
for all my machines and then I can monitor them all in one workspace of my (Windows+cygwin) desktop. It's getting difficult to monitor those 20-CPU Xeons with only a 1920x1200 monitor, tho'but! :-/
18) Message boards : Number crunching : Panic Mode On (91) Server Problems? (Message 1593587)
Posted 28 Oct 2014 by Profile ivan
It seems to take a long time for the database engine to restart - whether it's loading tables into memory, integrity-checking the indexes, or something else, I don't know. But after a few hours, it lets out an almighty sneeze, and suddenly everything works at full speed again.

Yeah, response is suddenly back to normal now, database queries are up, and reported delays are more like usual.
19) Message boards : Number crunching : Panic Mode On (91) Server Problems? (Message 1593580)
Posted 28 Oct 2014 by Profile ivan
If you're only reporting or requesting work you get through (though I've yet to get any new work), but if you do both at the same time you get the HTTP error response.

Looks like I've managed to get one MB and ten APV7 tasks tonight (all AP tasks on my work laptop).

Other than that the SSP updates the page time yet the other numbers don't seem to be changing at all.

Seems to be some problems with the database, e.g. I get no response when I try to look at my Errors page to see if there are any new errors.
20) Message boards : Number crunching : Panic Mode On (91) Server Problems? (Message 1592535)
Posted 26 Oct 2014 by Profile ivan
Cricket just took a huge dive. Expect snow, hail, and thunderstorms.

(Link added.)


Next 20

Copyright © 2014 University of California