Teburin Abubuwan Ciki
1 Gabatarwa
PUNCH4NFDI wata ƙungiya ce ta kimanin masana kimiyya 9,000 daga fannonin barbashi, taurari, barbashin taurari, hadron da ilimin kimiyyar nukiliya a cikin Jamus. Tare da tallafin Gidauniyar Bincike ta Jamus (DFG) a matsayin wani ɓangare na shirin Tsarin Bayanan Bincike na Ƙasa (NFDI), ƙungiyar tana da niyyar ƙirƙirar dandalin bayanan kimiyya na haɗin gwiwa wanda ke ba da damar FAIR (Ana iya Ganewa, Samun dama, Haɗin kai, Sake amfani da su) ga bayanai da albarkatun kwamfuta a duk cibiyoyin da ke shiga.
9,000+
Masana Kimiyya da aka wakilta
Shekaru 5
Lokacin Ba da Kudade na Farko
Da yawa
Al'ummomin Bincike
2 Tsarin Hadin Kan Kwamfuta iri-iri
Yunƙurin Compute4PUNCH yana magance ƙalubalen haɗa albarkatun kwamfuta daban-daban ciki har da Lissafi Mai Yawan Aiki (HTC), Lissafi Mai Ƙarfin Aiki (HPC), da albarkatun Girgije da cibiyoyin da ke shiga suka bayar a matsayin gudunmawa.
2.1 Tsarin Haɗin Albarkatu
Tsarin yana amfani da HTCondor a matsayin tsarin batch na sama, yana haɗa albarkatu iri-iri ta hanyar COBalD/TARDIS mai tsara albarkatu. Wannan hanyar tana ba da damar raba albarkatu a sarari yayin kiyaye tsarin aiki na yau da kullun a wuraren masu bayarwa.
2.2 Tsarin Shiga da Tabbatar da Hanyoyin Shiga
Tsarin Tabbatar da Hanyoyin Shiga da Ba da Izini (AAI) na tushen alama yana ba da daidaitaccen damar zuwa albarkatun lissafi. Ƙofofin shiga na al'ada da JupyterHub suna aiki a matsayin hanyoyin shiga, suna ba masu amfani madaidaitan hanyoyin haɗi ga tsarin haɗin gwiwar.
2.3 Gudanar da Yanayin Software
Fasahohin kwantena da Tsarin Fayil na Na'ura ta CERN (CVMFS) suna tabbatar da samar da yanayin software na musamman ga al'umma a cikin tsarin iri-iri.
3 Tsarin Hadin Kan Ajiya
Storage4PUNCH yana mai da hankali kan haɗa tsarin ajiya da al'umma ke bayarwa waɗanda suka fi dogaro da fasahar dCache da XRootD, suna amfani da hanyoyin da aka kafa sosai a cikin al'umar Kimiyyar Ƙarfin Lantarki (HEP).
3.1 Haɗin Fasahar Ajiya
Tsarin yana haɗa tsare-tsaren ajiya daban-daban ta hanyar daidaitattun ka'idoji da hanyoyin haɗi, yana ba da damar samun bayanai ɗaya a duk cibiyoyin da ke shiga yayin kiyaye yancin kai na gida.
3.2 Maganin Metadata da Ƙwaƙwalwar Ajiya
Ana kimanta fasahohin da ake da su don ajiya a cikin ƙwaƙwalwar ajiya da sarrafa metadata don haɗin kai mai zurfi, da nufin inganta gano bayanai da aikin samun dama a duk faɗin tsarin ajiya na haɗin gwiwa.
Bincike Mai mahimmanci: Kimanta Tsarin Hadin Kai
Gano Asali
Hanyar haɗin gwiwar PUNCH4NFDI tana wakiltar sasantawa mai ma'ana tsakanin raba albarkatu mai kyau da ƙuntatawa na aiki na tsarin da ake da shi. Tsarin ya yarda cewa a cikin lissafin kimiyya, shingen siyasa da ƙungiya sau da yawa sun fi ƙalubalen fasaha. Ta hanyar ginawa akan ingantattun fasahohi kamar HTCondor da dCache, suna yin aminci maimakon juyin juya hali.
Tsarin Ma'ana
Ci gaban fasaha yana bin tsari bayyananne: fara da abin da ke aiki (kafaffen kayan aikin HEP), ƙara yadudduka na haɗin gwiwa (COBalD/TARDIS), da rage rushewar ayyukan da ake da su. Wannan hanya ta ƙara ya bambanta sosai da ƙarin ƙwaƙƙwaran ayyukan lissafin grid kamar Tsarin Grid na Turai (EGI) waɗanda sau da yawa sukan yi fama da karɓuwa saboda rikitarwa. AAI na tushen alama yana nuna koyo daga ƙalubalen sarrafa ainihin haɗin gwiwar da aka samu a cikin ayyuka kamar EduGAIN.
Ƙarfi & Kurakurai
Ƙarfi: Bukatar ƙaramin tsangwama ga masu ba da albarkatu yana da hazaka—yana rage shingen shiga sosai. Yin amfani da kwantena da CVMFS don rarraba software yana magance ɗaya daga cikin matsalolin da suka dore a cikin yanayin lissafi iri-iri. Mayar da hankali kan ingantattun fasahohin HEP yana ba da aminci nan take a cikin al'ummomin da suke niyya.
Kurakurai: Dogaro mai yawa akan HTCondor yana haifar da madaidaicin dogaro na gine-gine. Ko da yake an tabbatar da shi a cikin mahallin HEP, wannan hanyar na iya iyakance sassauci ga ayyukan da ba na HEP ba. Takardar ta bayyana kadan game da garanti na ingancin sabis ko hanyoyin fifita albarkatu—mahimman gibin don ayyukan kimiyya na samarwa. Idan aka kwatanta da ƙarin hanyoyin zamani kamar haɗin gwiwar tushen Kubernetes (kamar yadda aka gani a cikin aikin Science Mesh), tsarin gine-ginensu yana jin ɗan tsufa.
Abubuwan Fahimta masu Aiki
Ya kamata ƙungiyoyin bincike su yi koyi da hanyar PUNCH4NFDI ta-farko-masu-bayarwa amma su ƙara ta da ingantaccen manufofin matakin sabis. Ya kamata yadudduka na haɗin gwiwa su riƙaya zuwa fasahohin asalin girgije yayin kiyaye dacewar HTCondor. Mafi mahimmanci, dole ne su magance gibin haɗin gwiwar metadata—ba tare da ingantaccen sarrafa metadata na tsarin-tsare-tsare ba, ana iya gano bayanai a cikin haɗin gwiwar zai kasance mai iyaka. Duba ingantattun aiwatarwa kamar kayan gine-ginen Girgije na iya ba da darussa masu mahimmanci a cikin daidaita haɗin gwiwa da aiki.
4 Tsarin Bincike na Fasaha
Ana iya ƙirƙira matsalar raba albarkatu a cikin yanayi na haɗin gwiwa ta amfani da ka'idar ingantawa. Bari $R = \{r_1, r_2, ..., r_n\}$ ya wakilci tarin albarkatun da ake da su, kowanne yana da iyawa $C_i$ da amfani na yanzu $U_i$. Manufar ingantawa don rarraba aikin ana iya bayyana shi azaman:
$$\min\sum_{i=1}^{n} \left( \frac{U_i + w_j}{C_i} \right)^2 + \lambda\sum_{i=1}^{n} \sum_{j=1}^{m} d_{ij}x_{ij}$$
inda $w_j$ ke wakiltar aikin da ke shigowa $j$, $d_{ij}$ shine farashin canja wurin bayanai, kuma $x_{ij}$ shine ma'aunin yanke shawara na raba. Wannan aikin farashi na quadratic yana taimaka wajen daidaita kaya a cikin albarkatu iri-iri yayin rage yawan motsin bayanai.
Misalin Tsarin Bincike
Matrix na Yanke Shawarar Zaɓin Albarkatu:
Don aikin bincike na bayanan taurari na yau da kullun wanda ke buƙatar sa'o'in CPU 1000 da ajiyar wucin gadi na 5TB, tsarin yana kimanta:
- Albarkatun HTC: Mafi dacewa don ayyuka masu kama da juna, babban aikin aiki
- Albarkatun HPC: Ya dace da kwaikwayayen da aka haɗa, ƙarancin buƙatun jinkiri
- Albarkatun Girgije: Mai sassauci don ƙarfin fashewa, mafi tsada kowace sa'ar lissafi
Algorithm ɗin yanke shawara yana auna abubuwa ciki har da wurin bayanai, lokutan jira, da dacewar gine-gine don tura ayyukan aiki zuwa albarkatun da suka dace ta atomatik.
5 Sakamakon Gwaji da Aiki
Aiwatar da samfuran farko sun nuna yuwuwar hanyar haɗin gwiwar. Gwaji tare da aikace-aikacen kimiyya daga al'ummomin da ke shiga ya nuna:
- Nasarar gabatar da aiki a cikin masu ba da albarkatu daban-daban 5 ta amfani da takaddun shaida ɗaya
- Matsakaicin jinkirin farawa na aiki na dakika 45 a cikin albarkatun haɗin gwiwa
- Turawa yanayin software ta hanyar CVMFS yana rage lokacin saiti daga sa'o'i zuwa mintuna
- Haɗin ajiya yana ba da damar samun bayanai ta hanyar shiga tsakanin wurare tare da aiki a cikin kashi 15% na damar gida
Halayen ayyukan sun yi daidai da tsammanin tsarin haɗin gwiwar, inda fa'idodin tarawa albarkatu dole ne a daidaita su da yawan aikin haɗin kai da motsin bayanai a cikin yankunan gudanarwa.
6 Ayyuka na Gaba da Ci gaba
Tsarin haɗin gwiwar yana buɗe hanyoyi masu ban sha'awa da yawa don ci gaba na gaba:
- Ayyukan Koyon Injin: Tsawaita goyon baya ga albarkatu masu wadatar GPU da kwantena na tsarin ML
- Bincike Mai Ma'amala: Haɓaka haɗin JupyterHub don binciken bayanai na ainihi a cikin tarin bayanai na haɗin gwiwa
- Haɗin Kai na Ƙasashen Duniya: Yuwuwar haɗin kai tare da irin wannan kayan aiki a wasu ƙasashe biyo bayan tsarin lissafin LHC
- Haɗin Kan Kwamfutar Ƙididdiga: Shirye-shiryen ayyukan aiki na gargajiya-ƙididdiga masu haɗaka yayin da albarkatun ƙididdiga suka samu
Zanen tsarin yana ba da damar ƙara karɓar sabbin fasahohi yayin kiyaye dacewa ta baya tare da ayyukan aiki na yanzu.
7 Nassoshi
- Thain, D., Tannenbaum, T., & Livny, M. (2005). Distributed computing in practice: The Condor experience. Concurrency and Computation: Practice and Experience, 17(2-4), 323-356.
- Blomer, J., et al. (2011). Scaling CVMFS to many millions of files. Journal of Physics: Conference Series, 331(4), 042003.
- Frey, J., et al. (2002). Condor-G: A computation management agent for multi-institutional grids. Cluster Computing, 5(3), 237-246.
- European Grid Infrastructure. (2023). EGI Federated Cloud. Retrieved from https://www.egi.eu/federated-cloud/
- Science Mesh. (2023). Federated infrastructure for scientific collaboration. Retrieved from https://sciencemesh.io/
- Materials Cloud. (2023). A platform for open science in materials research. Retrieved from https://www.materialscloud.org/