Auger Xtrem Web Monte Carlo computation on A

Auger & Xtrem. Web: Monte Carlo computation on A Global Computing platform O. Lodygensky, G. Fedak, V. Neri, A. Cordier, F. Cappello Laboratoire de l’Accelerateur Lineaire; Laboratoire de Recherche en Informatique; CNRS, Université Paris sud, France. March 27, 2003 CHEP 2003 - O. Lodygensky

Sommaire • Introduction • Xtrem. Web • Auger distributed computing • Conclusion March 27, 2003 CHEP 2003 - O. Lodygensky 2

Different GRID « GRID » 2 distributed system types Global computing systems « Desktop GRID » « Internet Computing » Nodes caracteristics Traditionnal computing centers, Clusters Windows, Linux, Mac OS • <100 • Stables • Individually identified • Trusted • ~100 000 • Volatiles • No individual ident • Not trusted Peer to Peer systems (P 2 P) March 27, 2003 CHEP 2003 - O. Lodygensky 3

Desktop GRID • Dedicated applications One server centralizes scheduling On volunteer PCs • Production projects Client Application set params. / get results. – Folding@home, Genome@home, – Folderol, server parametres Internet Volunteer PC : load & exec task March 27, 2003 – SETI@Home, distributed. net, – Decrypthon Volunteer PC • Open source/research projects – Xtrem. Web, BOINC, • Commercial platforms – Entropia, Datasynapse, – United Devices, Grid systems CHEP 2003 - O. Lodygensky 4

Desktop Grid characteristics • Scalability : up to 100 k, 1 M hosts • Heterogeneity : different hardwares, OSes • Volatility : unpredictable participant behaviour Seti@home, Napster, Kazaa, etc. : they work well despite volatility. • Perenity : developments and upgrades must be easy • Performances : Seti@home ~30 Tflops, Kazaa (1 M users : 100 kb/s, 1 Mb/s 100 Gb/s, 1 Tb/s? ). • Sécurity : ü Volunteer PCs and servers integrity ; ü Prevent application & results corruption ; ü Authentication. March 27, 2003 CHEP 2003 - O. Lodygensky 5

Sommaire • Introduction • Xtrem. Web • Auger distributed computing • Conclusion March 27, 2003 CHEP 2003 - O. Lodygensky 6

XW : Architecture • Centralized • Global Computing (Peer to Peer) • 3 entities : client/coordinator/worker Hierarchical Coordinator P 2 P Coordinator Global Computing coordinator (client) PC Client/worker Internet / LAN March 27, 2003 PC Client/Worker PC Worker CHEP 2003 - O. Lodygensky PC Worker 7

XW : Technology Server Java Communication protocol XML-RPC SSL Worker Client Java SQL Java JDBC Data Base Http Server PHP 3 -4 Installation GNU autotool Pre requisite for installation: database (Mysql), JAVA > jdk 1. 2. March 27, 2003 CHEP 2003 - O. Lodygensky 8

XW : Security ssh Coordinat. Worker Client March 27, 2003 Sandbox (SBLSM) CHEP 2003 - O. Lodygensky Loaded App 9

XW : fault tolerance model • Every entity is volatile by essence • Connectionless protocols => All entities are stand alone Client 2 Submit task Sync/Retrieve result e r ieve r t e c /R Sync /Subm it task Sync /Retr ieve res Coord. Worker 1 Sync / Get work Put resu lt Coord. ult March 27, 2003 Get work Put result CHEP 2003 - O. Lodygensky Worker 2 work t e G / Sync result t u P / c n Sy 10

Sommaire • Introduction • Xtrem. Web • Auger distributed computing • Conclusion March 27, 2003 CHEP 2003 - O. Lodygensky 11

Pierre Auger Observatory Understanding the origin of very high cosmic rays: • Aires: Air Showers Extended Simulation – Sequential, Monte Carlo. Time for a run: 5 to 10 hours Traditional Super Computing Centers Air shower parameter database (Lyon, France) Xtrem. Web Server CINES (Fr) air shower Estimated PC number ~ 5000 PC worker Fermi Lab (USA) March 27, 2003 Internet and LAN PC worker Aires CHEP 2003 - O. Lodygensky PC Client PC Worker 12

Auger-XW (AIRES): High Energy Physics Application : AIRES Icluster Grenoble PBS Madison Wisconsin Condor Deployment: • Coordinator at LRI • Madison: 700 workers U-psud network Internet Pentium III, Linux (500 MHz+933 MHz) Other Labs (Condor pool) LRI Condor Pool • Grenoble Icluster: 146 workers (733 Mhz), PBS • LRI: 100 workers Pentium III, Athlon, Linux lri. fr XW Coordinator (500 MHz, 733 MHz, 1. 5 GHz) XW Client (Condor pool) March 27, 2003 CHEP 2003 - O. Lodygensky 13

Auger-XW (AIRES): High Energy Physics March 27, 2003 CHEP 2003 - O. Lodygensky 14

Auger-XW (AIRES): High Energy Physics March 27, 2003 CHEP 2003 - O. Lodygensky 15

Auger-XW (AIRES): High Energy Physics March 27, 2003 CHEP 2003 - O. Lodygensky 16

Auger-XW (AIRES): High Energy Physics March 27, 2003 CHEP 2003 - O. Lodygensky 17

Auger-XW (AIRES): High Energy Physics March 27, 2003 CHEP 2003 - O. Lodygensky 18

Auger-XW (AIRES): High Energy Physics March 27, 2003 CHEP 2003 - O. Lodygensky 19

Auger-XW (AIRES): High Energy Physics March 27, 2003 CHEP 2003 - O. Lodygensky 20

Sommaire • Introduction • Xtrem. Web • Auger distributed computing • Conclusion March 27, 2003 CHEP 2003 - O. Lodygensky 21

Conclusion Xtrem. Web : a « desktop Grid » platform Fault tolerance. Xtrem. Web : « connectionless » + « restartable » Security : certificats + crypto + sandbox +… What we have learned so far with Xtrem. Web: – Deployment is critical – When they understand the computational power potential , – users rapidly ask for more resources!!! Xtrem. Web Auger: International Desktop GRID Condor pools with XW as global infrastructure Good performances (ratio 1: 60 with several small hosts than the reference) =>Schedulling is a lack of Xtrem. Web <= =>Strong need of results browsing tools <= March 27, 2003 CHEP 2003 - O. Lodygensky 22

Software • Xtrem. Web : www. Xtrem. Web. net – Since 2001 – Acual version : 1. 2. rc 0 March 27, 2003 CHEP 2003 - O. Lodygensky 23