Analysis of Results of the Rating of Volunteer Distributed Computing Projects

Transcription

1 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org Analyi of Reult of the Rating of Volunteer Ditributed Computing Project Vladimir N. Yakimet 1,2[ ] and Ilya I. Kurochkin 1[ ] 1 Intitute for Information Tranmiion Problem of Ruian Academy of Science, Mocow, Ruia 2 The Ruian Preidential Academy of National Economy and Public Adminitration, Mocow, Ruia iakimet@mail.ru, kurochkin@iitp.ru Abtract. Volunteer ditributed computing (VDC) i a fairly popular way of conducting large cientific experiment. The organization of computational experiment on a certain ubject implie the creation of a project of volunteer ditributed computing. In thi project, computing reource are provided by volunteer. The community of volunteer i about everal million people around the world. To increae the computing power of the volunteer ditributed computing project, technical method for increaing the efficiency of computation can be ued. However, no le important are method of attracting new volunteer and motivating thi virtual community to provide computing reource. The organizer of VDC project, a a rule, are expert in applied field, but not in the organization of volunteer ditributed computing. To ait the organizer of the VDC project author conducted a ociological tudy to determine the motivation of volunteer, created a multiparameter method and rating for evaluating variou VDC project. Thi article propoe a method for aeing the trength and weaknee of VDC project, baed on the approach. The reult of multiparameter evaluation and rating of project can help the organizer of the VDC project to increae the efficiency of computation, and the community of volunteer to provide a tool for comparing the variou VDC project. Keyword: Volunteer ditributed computing (VDC) BOINC The VDC project Cruncher Volunteer Evaluation index of the VDC project Characteritic for aeing the quality of VDC project. 1 Introduction The ue of ditributed computing ytem for high-performance computing i an alternative to calculation on upercomputer and other multiproceor computer ytem. Ditributed computing ytem or grid ytem have a number of feature, uch a heterogeneity of computing node, their geographical ditance, untable network topology and high probability of diconnection of a computing node or communication channel. But even with uch feature, the computing potential of the grid ytem can be huge becaue of the large number (hundred of thouand) of compute node. 893

2 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org 2 There are oftware platform for organizing ditributed computing, uch a HTCondor [1], Legion [2], BOINC [3]. At the moment, the mot common platform for organizing ditributed computing i BOINC (Berkeley Open Infratructure for Network Computing) [4]. Public grid ytem attract the computing power of volunteer. A a rule, the organizer of public grid ytem are cientific or educational organization. A public computing grid ytem i called a volunteer ditributed computing project. Volunteer ditributed computing project (VDC project) are rather heterogeneou, both in term of topic and in their organization. There are long-lived project that remain active and uccefully conduct experiment for almot 20 year (SETI@home, Folding@home, Eintein@home). There are project that were created for one large computational experiment or cientific problem, after everal year, they top working (POGS@home, Poem@home). Some project imultaneouly run everal independent computing experiment (SZTAKI, LHC@home). Developer of ome project can pay great attention to the deign of the project ite (POGS@home, Acoutic@home), and other project are limited to uing the tandard template of the BOINC platform (Climate@Home, Optima@home[11]). Some project are conidered only on the CPU, other ue CPU + GPU. Interaction with volunteer in project ha general principle[13]: Attracting computing reource to help the cientific team; Publication of experimental reult; Publication of cientific and popular cientific publication on the project webite; Interaction with volunteer; Organization of competition for volunteer within the project; Motivation of volunteer with virtual prize and certificate. However, for each VDC project there are pecific feature of interaction with the community of volunteer. Some project i popular, a it wa organized by cientit with a global name (LHC@home). Another project ha a clear and high goal - the earch for a cure for cancer (Roetta@home). If a media project ha been told about a project, thi can give an impetu to the project even in the medium term - attracting new volunteer for everal month to 1 year. The participation of the project in virtual port (Formula BOINC, BOINC Pentathlon) can alo attract ignificant computational reource to the project. The webite of a number of VDC project baed on BOINC platform [3] were tudied a well a information about them on the ite boinc.ru. The lit of VDC project include: SETI@home one of the firt project on the bai of which the BOINC platform wa developed. Thi project of a group of cientit from Berkeley Univerity for proceing data from radio telecope. Ateroid@home project i organized by the Atronomical Intitute of Charle Univerity, Czech Republic. Project aimed at determining the hape and parameter of the rotation of ateroid according to photometric obervation In the proce of project implementation, the public databae DAMIT i filled. 894

3 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org 3 POGS@home project i organized by the International Centre for Radio Atronomy Reearch, Autralia. Thi project aimed at building a multipectral atla (from near infrared to ultraviolet radiation), a well a at determining the rate of tar formation, the tellar ma of galaxie, the ditribution of dut and it ma in the galaxy and etc.; SAT@home project i organized by the Intitute for Sytem Dynamic and Control Theory of Siberian Branch of Ruian Academy of Science, Ruia. Project aociated with earching for olution to uch complex problem a inverion of dicrete function, dicrete optimization, bioinformatic, and etc., which can be effectively reduced to the problem of the feaibility of Boolean formula[9]; Roetta@home project i organized by Intitute for Protein Deign, Univerity of Wahington, USA. Project aociated with olving one of the bigget problem in molecular biology - the calculation of the tertiary tructure of protein from their amino acid equence; MilkyWay@home project i organized by Renelaer Polytechnic Intitute, USA. Thi project tudie the hitory of our galaxy by analyzing the tar in the Milky Way galaxy' Galactic Halo; Some other VDC project Folding@home[5], Eintein@home[6], LHC@home[7][10], Geraim@home[8]and etc. It hould be noted that in Englih and Ruian literature a majority of publication and article on Internet reource provide a decription of the individual VDC project [8][9][10]. Significantly le common are cientific paper which characterize organization of activitie and the involvement of cruncher [11][14]. And it i quite rare to ee work which examine variou apect of citizen participation in VDC project [12, 13]. One of the firt attempt of more or le ytematic tudy of the condition and characteritic of cruncher participating in VDC project in Ruia, a well a their motivation and preference ha become a ociological tudy among 650 Ruian cruncher [15]. 2 Decription of the Methodology for the Analyi of VDC Project The methodology for the index evaluation of VDC project include the following main tage: 1. Creating an index model; 2. Collect information to decribe the mot important element of VDC project. 3. Conducting urvey of participant of elected VDC project to calculate the value of the YaK-index and etimate. 4. Perform calculation and viualize the reult. The YaK-index wa developed to ae the quality of the VDC project implementation, which wa applied to a number of the above-mentioned project in order to identify their "weak" ide, develop propoal to increae their efficiency, increae 895

4 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org 4 attractivene for people intereted in the VDC, and providing comparable information for the organizer of VDC project. 2.1 Model of the YaK-index of the VDC Project We introduce the notation: i 1, n - the ordinal number of the important characteritic of the VDC project (hereinafter referred to a characteritic),it i aumed that n equal to 7-9, that i, from 7 to 9 etimated characteritic of each VDC project will be taken into account; 1, S - the ordinal number of the VDC project, S=34 (for thi paper); R - YaK-index of the VDC project ; xi - availability of characteritic i of the VDC project (0 if not preent;1 if available); K - number of repondent for the VDC project ; K - number of repondent for all VDC project; - the mean weighting factor (ignificance) of characteritic i of the VDC project i from i K repondent, 0 i 1, i 1 ; n i 1 - the mean expert aement of the quality of i-th characteritic of -th VDC project, The cale value vary from -2 to 2. A linguitic interpretation of thee value i given in the quetionnaire. If neceary i converted into the et. Identically for all n characteritic it i mapped one-to one in a numerical et from the et of poible linguitic etimate. The maximum value of the numeric cale i m. In our cae, it i aumed that m = 5. The index value vary from 1 to 5. But it i poible to normalize the index value o that they vary from 0 to 1. There are two poibilitie for calculating the value R : 1. When the weight of characteritic for a VDC project are individual and independent of what uch are uch weight for all other VDC project. 2. When for all VDC project the ame vector of weight i defined. In the firt cae, the index R i calculated a follow: R 1 n i 1 x i i i nm 1. (1) Here n 1 i the number of characteritic for VDC project, n1 In the econd cae n. 896

5 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org 5 R 2 n i 1 x i i i nm 1. (2) 2.2 Collect Information for Decribing the Mot Important Element of VDC Project For the urvey of volunteer, a preliminary lit of characteritic for project evaluation wa created. At conference and round table with repreentative of the volunteer community and expert, dicuion were held on thi lit and ome correction were made. A a reult, 9 characteritic were elected to ae the feature of VDC project[16]: viion cientific and cientific-popular publication on the topic 4. Deign of the project (ite, certificate, creenaver); material on the project ite; 6. Viualization of the project reult (photo, feedback (forum, chat room, etc.); the project (competition, coring ytem, prize); barrier and organizational or technical difficultie). The aement of each characteritic wa uppoed to be 5-point. The quetionnaire wa initially focued on the Ruian-peaking audience, o the following aement were propoed (with the following linguitic interpretation): 1. "+2" i excellent; 2. "+1" i good. 3. "0" i normal; 4. "-1" - it i neceary to improve; 5. "-2" i bad. Thi interpretation of the etimate allow you to get rid of the unneceary connotation with chool grade (from 1 to 5) and make it poible to ue the entire range of etimate. Due to the heterogeneity of the project, each characteritic ha a different impact on the evaluation of the project. Therefore, it wa decided to ae the ignificance of each characteritic for a particular project. Weight characteritic could vary from "0" - not ignificant, to "10" - very ignificant. In addition to evaluating the project, everal quetion were aked in the quetionnaire on the degree of involvement of the 897

6 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org 6 repondent in thi project: determining the repondent' tatu, determining the duration of the repondent' interaction with the project. 2.3 Conducting Survey of Participant of VDC Project The quetionnaire wa implemented uing Google Doc in Ruian and Englih [21]. Information about the quetionnaire wa ditributed by the adminitration of the ite BOINC.ru on the profile forum and ite of the community of volunteer. A a reult, etimate were received from 259 repondent for 34 project. However, for ubequent proceing, only thoe project for which more than 10 quetionnaire were filled were taken (Fig. 1). Total number SAT@home SETI@home Eintein@home Roetta@home Geraim@home POGS@home Ateroid@home LHC@home MilkyWay@home Folding@home Number of repondent Fig. 1. Ditribution of the number of repondent among project. The reult were obtained in , therefore new project uch a RakeSearch, XANSONS for COD, etc., are not lited in the rating. 2.4 Viualization of Survey Reult To determine the averaged weight of each characteritic, quetionnaire were ued from all 259 repondent (for all project). When comparing the average weight of a characteritic for individual project, the value were not ignificantly different from the average value for individual project. A an example, the project characteritic are ranked by weight according to the average value for all project (Fig. 2). 898

7 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org viion of the the project cientific and feedback (forum, material on the 6. Viualization of the project reult 4. Deign of the project (ite, cruncher All project Fig. 2. Ranking of characteritic by weight. For the viual preentation of the project characteritic a radar diagram wa ued (Fig. 3). The characteritic are arranged in decending order of average weight. All project 9 the project (competition, viion coring ytem, prize); 8 4. Deign of the project (ite, certificate, creenaver); 7 6 barrier and 6. Viualization of the project reult (photo, material on the project ite; feedback (forum, chat room, etc.); cientific and cientificpopular publication on the topic Fig. 3. Ranked weight of characteritic in radar diagram The average etimation for all project on the radar diagram (Fig.4) are given, while the ranking of the characteritic in decending order of importance i preerved (Fig. 2 and Fig.3). 899

8 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org 8 the project (competition, coring ytem, prize); 4. Deign of the project (ite, certificate, creenaver); 2,0 1,5 1,0 0,5 0,0 All project viion barrier and 6. Viualization of the project reult (photo, material on the project ite; feedback (forum, chat room, etc.); cientific and cientificpopular publication on the topic Fig. 4. Etimate of the characteritic of all project, taking into account their ranking by weight 3 Reult for Selected VDC Project Let' preent etimate of large international VDC project: Folding@home, Eintein@home, LHC@home and Ruian VDC project Geraim@home. 3.1 Folding@home Folding@home i a ditributed computing project for dieae reearch that imulate protein folding, computational drug deign, and other type of molecular dynamic [10]. The project tarted in 2000 and i one of the mot ucceful international project that began to ue grid ytem from peronal computer. The audience of the project i more than 1 million uer and continue to grow. For everal year, the organizer of Folding@home have maintained compatibility with the BOINC client. Thi allowed the volunteer, who ue a BOINC client, to participate in the project. But, depite thi, the community of volunteer i very different from the volunteer community of VDC project on the BOINC platform. Thi can be een from the variou etimate of the ignificance of the characteritic (Fig.5). The ucce and attractivene of the project for volunteer can be een at Fig.6, Where project etimate for each characteritic are better than average value for all project. 900

9 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org 9 the project (competition, coring 4. Deign of the project (ite, certificate, creenaver); 10,00 9,00 8,00 7,00 6,00 5,00 All project Folding@home viion barrier and 6. Viualization of the project reult (photo, material on the project ite; feedback (forum, chat room, etc.); cientific and cientificpopular publication on the topic Fig. 5. Weight of characteritic for the project SETI@home the project (competition, coring 4. Deign of the project (ite, certificate, creenaver); 2,00 1,50 1,00 0,50 0,00 All project Folding@home viion barrier and 6. Viualization of the project reult (photo, material on the project ite; feedback (forum, chat room, etc.); cientific and cientificpopular publication on the topic Fig. 6. Evaluation of the characteritic of the project Folding@home 3.2 Eintein@home Eintein@home project i upported by the American Phyical Society, the US National Science Foundation, the Max Planck Society. Eintein@home organized to 901

10 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org 10 earch for weak atrophyical ignal from pinning neutron tar uing data from the LIGO gravitational-wave detector, the Arecibo radio telecope, and the Fermi gamma-ray atellite [11]. The project i one of the larget international VDC project on the BOINC platform. The project i active now. About 500 thouand uer were connected to the project. In fact, the Eintein@home project i a tandard project, a the evaluation of the ignificance of the characteritic weight coincide with the average value for all project (Fig.7) and the project etimate by characteritic are higher than the value of the averaged weight for all project except for one - timulation of the participation of the cruncher in the project (Fig.8). the project (competition, coring 4. Deign of the project (ite, certificate, creenaver); 10,00 9,00 8,00 7,00 6,00 5,00 All project Eintein@home viion barrier and 6. Viualization of the project reult (photo, material on the project ite; feedback (forum, chat room, etc.); cientific and cientificpopular publication on the topic Fig. 7. Weight of characteritic for the project Eintein@home 902

11 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org 11 the project (competition, coring 4. Deign of the project (ite, certificate, creenaver); 2,00 1,50 1,00 0,50 0,00 All project viion barrier and 6. Viualization of the project reult (photo, material on the project ite; feedback (forum, chat room, etc.); cientific and cientificpopular publication on the topic Fig. 8. Evaluation of the characteritic of the project 3.3 project help phyicit from CERN compare theory with experiment, in the earch for new fundamental particle and anwer to quetion about the Univere [12]. In thi project, everal different computing tak are being olved. The project ha a large audience of about 160,000 volunteer. At preent the project uccefully function and develop. The cale and importance of the cientific problem being olved even reduce the requirement of the volunteer community for uch a characteritic a the timulation of the participation of the cruher in the project (Fig.9). Mot of hi etimate are above the average for all project (Fig.10). 903

12 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org 12 the project (competition, coring ytem, prize); 4. Deign of the project (ite, certificate, creenaver); 10,00 9,00 8,00 7,00 6,00 5,00 All project viion barrier and 6. Viualization of the project reult (photo, material on the project ite; cientific and cientific-popular publication on the topic feedback (forum, chat room, etc.); Fig. 9. Weight of characteritic for the project the project (competition, coring 4. Deign of the project (ite, certificate, creenaver); 2,00 1,50 1,00 0,50 0,00 All project viion barrier and 6. Viualization of the project reult (photo, material on the project ite; feedback (forum, chat room, etc.); cientific and cientificpopular publication on the topic Fig. 10. Evaluation of the characteritic of the project 904

13 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org The Ruian project i notable for the fact that the erver part of the project i implemented for the MS Window operating ytem, but the tandard client part of the BOINC platform i ued [13]. Number of uer about It i worth noting ignificant difference in the weight of characteritic in the project Geraim@home and for all project (Fig.11). A ignificant failure in ubmitting publication for the volunteer community and feedback problem are clearly viible in (Fig.12) And may be the project' primary improvement for project organizer. All project Geraim@home the project (competition, coring 4. Deign of the project (ite, certificate, creenaver); 10,00 9,00 8,00 7,00 6,00 5,00 viion barrier and 6. Viualization of the project reult (photo, material on the project ite; feedback (forum, chat room, etc.); cientific and cientificpopular publication on the topic Fig. 11. Weight of characteritic for the project Geraim@home 905

14 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org 14 the project (competition, coring 4. Deign of the project (ite, certificate, creenaver); 2,00 1,50 1,00 0,50 0,00 All project viion barrier and 6. Viualization of the project reult (photo, material on the project ite; feedback (forum, chat room, etc.); cientific and cientificpopular publication on the topic Fig. 12. Evaluation of the characteritic of the project 4 Performance of Calculation of the YaK-index for Selected VDC Project Uing the calculated average (for the conidered ubet of project) and individual value of the weight of the characteritic of the VDC project for each project eparately, a well a expert etimate of the characteritic, the value of the YaK-index were calculated (Table 1). Table 1. YaK-index. Project Title YaK-index (with average weight of characteritic) YaK-index (with individual weight of characteritic) 1 POGS@home Eintein@home Folding@home LHC@home Ateroid@home Roetta@home SETI@home MilkyWay@home Geraim@home SAT@home

15 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org 15 Judging by the magnitude of the YaK-index value, all 10 VDC project have a certain capabilitie for development. So, in the cae of uing individual characteritic cale (right column), the highet value of the index (0.69) have both the Eintein@home project, and the POGS@home project, and the lowet value ha the SAT@home project. The team of the firt two project, referring to the value of the characteritic weight, can determine which characteritic of their project they hould pay attention to in order to increae the value of the YaK-index. For the project Eintein@home, judging by Fig. 7 and Fig. 8, the growth of the value of the YaK-index i poible when a viualization of the project reult and a timulation of the participation of the cruncher in the project are improved. The lowet value of the YaK-index in both cae (right and econd from the right column of the table) were obtained by the Geraim@home and SAT@home project. Thi i jut a bit over half of the maximum poible value of the index. It i clear that thi project i "younger" than the ret one. Neverthele, by comparing the radar diagram with weight and etimate of characteritic, we can recommend to the project team to pay attention to the cientific component and decription of the project' concept, to viualize the reult and deign the project ite. 5 Concluion The propoed approach to aeing the tate of VDC project through a pecially organized urvey of project participant and calculation of the value of the developed YаK-index allow team to obtain in viualized form comparable information on 9 ignificant characteritic. Thee material can be an important tool for project team to improve work and project management. The etimated value for a number of well-known VDC project how the adequacy of the propoed approach, clearly illutrate the advantage of well-organized project with a olid operating experience, and create certain condition for monitoring the proce of formation and development of newly created and "young" VDC project. Acknowledgement Thi work wa funded by Ruian Science Foundation ( ). Reference 1.M.J. Litzkow, M. Livny, M.W. Mutka Condor-a hunter of idle worktation, Ditributed Computing Sytem, IEEE, Grimhaw, A. S., & Wulf, W. A. (1997). The Legion viion of a worldwide virtual computer. Communication of the ACM, 40(1), D.P. Anderon BOINC: a ytem for public-reource computing and torage, Grid Computing, IEEE,

16 Суперкомпьютерные дни в России 2018 // Ruian Supercomputing Day 2018 // RuianSCDay.org 16 4.The erver of tatitic of volunteer ditributed computing project on the BOINC platform. 5.Homepage of the project (date of acce: ). 6.Homepage of the project (date of acce: ). 7.Homepage of the project (date of acce: ). 8.Vatutin E.I., Titov V.S. Volunteer ditributed computing for olving dicrete combinatorial optimization problem uing project // Ditributed computing and gridtechnologie in cience and education: book of abtract of the 6th international conference. Dubna: JINR, Poypkin, M., Semenov, A., & Zaikin, O. (2012). Uing BOINC dektop grid to olve large cale SAT problem. Computer Science, 13(1), 25. (date of acce: ). 10.Lombraña González, D., Harutyunyan, A., Segal, B., Zacharov, I., McIntoh, E., Jone, P. L.,... & Grey, F. (2012). A volunteer computing ytem for maive numerical imulation of beam dynamic and high energy phyic event. In Conf. Proc. (Vol , No. IPAC-2012-MOPPD061, pp ). 11.Zaikin, O.S., Poypkin, M.A., Semenov, A.A., Khrapov, N.P.: Experience in organizing volunteer computing: a cae tudy of the OPTIMA@home and SAT@home project. Vetnik of Lobachevky State Univerity of Nizhni Novgorod, No. 5 2, pp (2012). 12.V. I. Tihchenko, A. L. Prochko. Ruian participant in BOINC-baed volunteer computing project. The activity tatitic. Computer Reearch and Modeling, 2015, vol. 7, no. 3, pp Clary EG, Snyder M, Ridge RD, Copeland J, Stuka AA, Haugen J, Miene P. Undertanding and aeing the motivation of volunteer: a functional approach. Journal of peronality and ocial pychology Jun;74(6): Webpage of World Community Grid project Member Study: Finding and Next Step ource= &utm_medium= &utm_campaign=uer_tudy_ (date of acce: ). 15.Yakimet V.N., Kurochkin I.I.. Voluntary ditributed computing in Ruia: a ociological analyi. Collection of cientific article of the XVIII Joint Conference "Internet and Contemporary Society" (IMS-2015), St. Peterburg, June 23, 2015, St. Peterburg: ITMO Univerity, Sc ISBN Kurochkin I.I., Yakimet V.N. Evaluation of the voluntary ditributed computing project SAT@home. National Supercomputer Forum (NSCF-2016), Perelavl-Zaleky, Ruia, November 29 - December 2,