Náročnost projektů - souhrná tabulka
Moderátoři: zdespi, Moderátoři
Dalsi upgrade.
Doplneny casy pro CPDN Sulphur a Slab pro Intel (na nove verzi 4.22) + dalsi drobnosti.
Pridal jsem dalsi radek - SETI Enhanced; jsou to predbezna cisla, ale napriklad naroky na pamet atp. jsou jiste vyssi nez u oficialni stare verze nebo jeji optimalizovane podobe.
Doplneny casy pro CPDN Sulphur a Slab pro Intel (na nove verzi 4.22) + dalsi drobnosti.
Pridal jsem dalsi radek - SETI Enhanced; jsou to predbezna cisla, ale napriklad naroky na pamet atp. jsou jiste vyssi nez u oficialni stare verze nebo jeji optimalizovane podobe.
Naposledy upravil(a) Honza dne čtv úno 16, 2006 3:39 am, celkem upraveno 1 x.
Jsem slepej, nebo nevidim?Honza píše:Dalsi upgrade.
Pridal jsem dalsi radek - SETI Enhanced;
A jeste jedna pripominka, Albert mi odesila cca. 155kB.
Private (old) + CCU stats.
-
LiborA
- BOINC Guru
- Příspěvky: 1298
- Registrován: čtv pro 08, 2005 10:11 am
- Bydliště: Chomutov, Praha, Moldava
- Kontaktovat uživatele:
Albert 4.37
Poslední dobou mi Albert 4.37 hlásí dobu výpočtu něco přes 48 minut a reálně to stíhá za 45 minut 
zatímco na druhém PC hlásí 6 hodin 12 minut 
Obojí jsou Intelovské mašiny, ta rychlá je Xeon 2,4 GHz, ta pomalá Mobilní Pentium na 1,6 GHz. Ty rozdíly jsou dost divné.
Doplněno: kredit na té rychlé mašině je pouze okolo 6 (5,79 až 7,1 ale přes 7 jsou dostal pouze 2 x z asi 100 výpočtů). Na té pomalé mašině jsou kredity 45 do 53
zatímco na druhém PC hlásí 6 hodin 12 minut Obojí jsou Intelovské mašiny, ta rychlá je Xeon 2,4 GHz, ta pomalá Mobilní Pentium na 1,6 GHz. Ty rozdíly jsou dost divné.
Doplněno: kredit na té rychlé mašině je pouze okolo 6 (5,79 až 7,1 ale přes 7 jsou dostal pouze 2 x z asi 100 výpočtů). Na té pomalé mašině jsou kredity 45 do 53
Ahoj vespolek,
dole je predelana a aktualizovana tabulka; trochu jsem ji optimalizoval. Zatim jsem nemel cas k ni napsat popis cesky, tak zatim s timhle...
_______
Below is an updated and optimalized version of comparation table you may have seen here and elsewhere. Over the period, I have been given some suggestions (which some of them were incorporated, other denied) and also spot some confusions. To avoid later, I'm giving some notes and explanations of the table.
Feel free to post suggestions, corrections and updates so that the table is up-to-date and gives a real value.
GENERAL NOTE:
All figures are for Windows XP OS. Some values are likely to differ on Linux or MacOS, mainly RAM usage, application size (first download), application version or even time-to-complete and/or credit.
COLUMNS EXPLANATION:
Project name - may include subprojects (CPDN Slab models, Sulphur cycle model or Coupled model) or some versions of project (SETI official/unoptimalized, optimalized, Enhanced application).
no GUI - RAM usage with no GUI (graphic open Windows not opened). It doesn't says if application has GUI or doesn't; albeit most projects do.
Internet traffic
Some users have slow (dial-up) or limited bandwidth internet connection; internet traffic generated by BOINC projects may be an issue. 3 columns try to delineate demands of each project. Note that when running several projects, you need to average each upon your resource share.
1st WU - download demands on first WU which includes: application + project files (CPDN model inputs, life-science protein database etc) and finally one WU. Usually, there is no need to re-download whole again unless (i) new application is released, (ii) new input files for WUs are deployed, (iii) project reset.
1 WU down/up - first number enumerate size of the WU for download (data for analysis); number after slash is the usual size of WU for upload (results size).
MB/month - when running project on 24/7 basis, 100% dedicated on dual-cored CPU (be it reference machine of AMD or Intel, whatever generates more results - see below), traffic may considerably grows and make it almost impossible to participate for some users. In other words - users provide idle CPU cycles and HD space (see next) but also internet bandwidth.
HD space
Each project needs it's application, downloaded WU, some temporary files and intermediate results and it occupies HD space. [CPDN Slab and SC models were compresed and stored on local disk for option of further analysis - see the second number after slash]. Total HD space depends on how many WU are you running (see max size) and how many WU are ready to crunch (see WU down).
Times to completed on WU
This is not a bechmark; it gives an average numbers on reference machines (usually those I possed to get the numbers among all projects). Also, numbers do change on other HW specification such as RAM speed, FSB, cache size etc. TO make numbers more accurate would only mean to make table several times larger.
AMD - reference machine is AMD64 3000+ or X2 3800+ running 2 WUs.
Intel - Pentium 4 at 3.GHz with HyperThreading on and running 2 WUs tu utilitize extra ~20% of performance.
Intel - Pentium D 820; a dual-cored CPU running 2 WUs.
Work Units (WU)
Deadline - time limit to report WU results since their download. Smaller numbers tends to give credit faster, longer periord may be suitable for dial-up/non-permanent internet connections. There are much more consequences with scheduler state, possible WU cache etc, but those topics are covered elsewhere.
Quota - each project has a maximum daily WU quota per CPU (limit on how many WU can be downloaded). This mechanism prevents trashing large portion of WUs on project server from unstable computers (or project application). Note, that this number decrease when host becomes less reliable (invalid WUs) and gets to it's maximum after issue is solved.
Quorum - most projects use a redundancy to ensure that results are reliable; quorum is the number of results need in order to validate WU. Until this number is reached, credit is pending; it is granted after validation on server side.
Application/WU version
Most project are in constant development and new version of application may occur. This may change other parameters of project, such as memory usage, aplication size, times to complete etc.
Credit - do I need to say something there? Science matters
___________________________________________________
BOINC Synergy thanks Honza for creating this useful table!
___________________________________________________
dole je predelana a aktualizovana tabulka; trochu jsem ji optimalizoval. Zatim jsem nemel cas k ni napsat popis cesky, tak zatim s timhle...
_______
Below is an updated and optimalized version of comparation table you may have seen here and elsewhere. Over the period, I have been given some suggestions (which some of them were incorporated, other denied) and also spot some confusions. To avoid later, I'm giving some notes and explanations of the table.
Feel free to post suggestions, corrections and updates so that the table is up-to-date and gives a real value.
GENERAL NOTE:
All figures are for Windows XP OS. Some values are likely to differ on Linux or MacOS, mainly RAM usage, application size (first download), application version or even time-to-complete and/or credit.
COLUMNS EXPLANATION:
Project name - may include subprojects (CPDN Slab models, Sulphur cycle model or Coupled model) or some versions of project (SETI official/unoptimalized, optimalized, Enhanced application).
no GUI - RAM usage with no GUI (graphic open Windows not opened). It doesn't says if application has GUI or doesn't; albeit most projects do.
Internet traffic
Some users have slow (dial-up) or limited bandwidth internet connection; internet traffic generated by BOINC projects may be an issue. 3 columns try to delineate demands of each project. Note that when running several projects, you need to average each upon your resource share.
1st WU - download demands on first WU which includes: application + project files (CPDN model inputs, life-science protein database etc) and finally one WU. Usually, there is no need to re-download whole again unless (i) new application is released, (ii) new input files for WUs are deployed, (iii) project reset.
1 WU down/up - first number enumerate size of the WU for download (data for analysis); number after slash is the usual size of WU for upload (results size).
MB/month - when running project on 24/7 basis, 100% dedicated on dual-cored CPU (be it reference machine of AMD or Intel, whatever generates more results - see below), traffic may considerably grows and make it almost impossible to participate for some users. In other words - users provide idle CPU cycles and HD space (see next) but also internet bandwidth.
HD space
Each project needs it's application, downloaded WU, some temporary files and intermediate results and it occupies HD space. [CPDN Slab and SC models were compresed and stored on local disk for option of further analysis - see the second number after slash]. Total HD space depends on how many WU are you running (see max size) and how many WU are ready to crunch (see WU down).
Times to completed on WU
This is not a bechmark; it gives an average numbers on reference machines (usually those I possed to get the numbers among all projects). Also, numbers do change on other HW specification such as RAM speed, FSB, cache size etc. TO make numbers more accurate would only mean to make table several times larger.
AMD - reference machine is AMD64 3000+ or X2 3800+ running 2 WUs.
Intel - Pentium 4 at 3.GHz with HyperThreading on and running 2 WUs tu utilitize extra ~20% of performance.
Intel - Pentium D 820; a dual-cored CPU running 2 WUs.
Work Units (WU)
Deadline - time limit to report WU results since their download. Smaller numbers tends to give credit faster, longer periord may be suitable for dial-up/non-permanent internet connections. There are much more consequences with scheduler state, possible WU cache etc, but those topics are covered elsewhere.
Quota - each project has a maximum daily WU quota per CPU (limit on how many WU can be downloaded). This mechanism prevents trashing large portion of WUs on project server from unstable computers (or project application). Note, that this number decrease when host becomes less reliable (invalid WUs) and gets to it's maximum after issue is solved.
Quorum - most projects use a redundancy to ensure that results are reliable; quorum is the number of results need in order to validate WU. Until this number is reached, credit is pending; it is granted after validation on server side.
Application/WU version
Most project are in constant development and new version of application may occur. This may change other parameters of project, such as memory usage, aplication size, times to complete etc.
Credit - do I need to say something there? Science matters
___________________________________________________
BOINC Synergy thanks Honza for creating this useful table!
___________________________________________________
2 jardaM: Ad. 2) U Ensteina je to jinak nez treba u SETI. Jednotky ktere maji spolecnych prvnich 9 znaku nazvu se pocitaji ze stejneho souboru. Ten je samozrejme ulozen v adresari ./projects/einstein.phys.uwm.edu
PS: To je moje dedukce, tak mi nekamenujte, kdybych byl mimo...
2 Honza: Pokud mam pravdu, tak je traffic v tvy tabulce uplne mimo a muze byt od stovek kB do stovek MB mesicne.
PS: To je moje dedukce, tak mi nekamenujte, kdybych byl mimo...
2 Honza: Pokud mam pravdu, tak je traffic v tvy tabulce uplne mimo a muze byt od stovek kB do stovek MB mesicne.
Private (old) + CCU stats.
Jo, s Einsteinem je trochu problem urcit delku vypoctu a tudiz i tne blbej kredit.
S trafikem je to take opravku nesnadne a muze se to pohybovat ve slusnem rozpeti podle toho jak se to sejde.
U casu davam rozpeti, stejne tak u kreditu a klidne to muzeme upresnit; u traffiku to davam spise radove, aby byla predstava.
Jak psal dejv, ze pri otpim to dela 4x tolik, tak to je z hlediska down/up WU pravda, jenomzne to neresi tahani dat k analyze
S trafikem je to take opravku nesnadne a muze se to pohybovat ve slusnem rozpeti podle toho jak se to sejde.
U casu davam rozpeti, stejne tak u kreditu a klidne to muzeme upresnit; u traffiku to davam spise radove, aby byla predstava.
Jak psal dejv, ze pri otpim to dela 4x tolik, tak to je z hlediska down/up WU pravda, jenomzne to neresi tahani dat k analyze
K te otazce, co ma JardaM a co nebylo moc spravne zodpovezena. SETI je v tomto nejprimitivnejsi, resp. nejcitelnejsi: aplikace a ke kazde WU je datain a dataout. Vetsina ostatnich projektu to ma reseno vice soubory na jednu WU, parametry aplikaci, vice vystupnich souboru, prave s nimi (zapokava/rozpakovani) atd.
VSECHNY BOINC projekty maji DATA i APLIKACI k analyze ve svem folderu, tj \project\jmeno_projektu. Ale zaroven VSECHNY WUs jsou zaznamenany v client_state.xml. To je zaroven nekdy problem urceni velikost trafiku a mista na disku - jako zde v pripade Einsteina.
Takze lze v dusledku tvrdit, ze ve folderu projektu nejsou zadne WU k pocitani, ale pouze aplikace a jeji data; samotne WU, resp. parametry aplikaci, ktere je pocitaji jsou v client_state.xml.
Ano, ve slots jsou pouze udaje pro rozpocitane WUs (ale nelze jiz tvrdit, ze tam jsou vzdy i rozpocitana data).
VSECHNY BOINC projekty maji DATA i APLIKACI k analyze ve svem folderu, tj \project\jmeno_projektu. Ale zaroven VSECHNY WUs jsou zaznamenany v client_state.xml. To je zaroven nekdy problem urceni velikost trafiku a mista na disku - jako zde v pripade Einsteina.
Takze lze v dusledku tvrdit, ze ve folderu projektu nejsou zadne WU k pocitani, ale pouze aplikace a jeji data; samotne WU, resp. parametry aplikaci, ktere je pocitaji jsou v client_state.xml.
Ano, ve slots jsou pouze udaje pro rozpocitane WUs (ale nelze jiz tvrdit, ze tam jsou vzdy i rozpocitana data).
Od minula, co jsem psal do tohoto threadu, ubehlo asi 15 updatu...tak jen ty posledni veci:
- update aplikace Rosetty
- update aplikace SETI Enhanced
- updaty aplikace Primegen
Beatminister me upozornil na to, ze za optimalizovanou verzi je vlastne stejne kreditu jako za oficialni pomalou (Einstein), podobne jako to mam u projektu s sumem, takze jsou to opravil. Moholo u nekterych lidi budit dojem, ze za optimalizovanou dostavaji na hodinu mene kreditu...a pritom opat je pravdou.
Jestli tam vidite nejake zastaralosti nebo primo blbosti, upozornete...dik
- update aplikace Rosetty
- update aplikace SETI Enhanced
- updaty aplikace Primegen
Beatminister me upozornil na to, ze za optimalizovanou verzi je vlastne stejne kreditu jako za oficialni pomalou (Einstein), podobne jako to mam u projektu s sumem, takze jsou to opravil. Moholo u nekterych lidi budit dojem, ze za optimalizovanou dostavaji na hodinu mene kreditu...a pritom opat je pravdou.
Jestli tam vidite nejake zastaralosti nebo primo blbosti, upozornete...dik
Naposledy upravil(a) Honza dne ned čer 18, 2006 9:40 am, celkem upraveno 1 x.
Má pravdu, finta je v něčem jiném. Protože můj kredit je nejnižší, škrtá se a bere se (i pro mně) hodnota prostřední, tedy pokud ten druhý a třetí neoptimalizuje taky. Ale já jsem na tu výslednou hodnotu spotřeboval polovinu, třetinu, čtvrtinu času, takže za stejný čas spočtu dvakrát, třikrát,čtyřikrát tolik.Honza píše:Beatminister me upozornil na to, ze za optimalizovanou verzi je vlastne stejne kreditu jako za oficialni pomalou (Einstein),