HPCI 1 2 3 4 5 1, 6 5 24 HPCI HPC OS HPC RENKEI-PoP Design of Advanced Software Deployment Infrastructure in HPCI Wide-area Distributed Environment Shinichiro Takizawa, 1 Masaharu Munetomo, 2 Atsuya Uno, 3 Taizo Kobayashi, 4 Hideyuki Jitsumoto, 5 Satoshi Matsuoka 1, 6 and Yutaka Ishikawa 5 The purpose of HPCI, which will be operated from autumn 2012, is to support HPC researchers to use K supercomputer, and its initial services are a federated authentication and global file sharing between K and supercomputers provided by computer centers in Japan. However, supercomputers are not suitable for HPC system researchers as their operations do not give users enough privileges. We design the advanced software deployment infrastructure that hosts distributed systems where researchers can have administrator privileges. We introduce the design of the system and a precedent system implemented on RENKEI-PoPs that use the same software. 1. ( HPCI) HPCI (HPC) 24 HPCI OS HPC OS 1 Tokyo Institute of Technology 2 Hokkaido University 3 RIKEN Advanced Institute of Computational Science 4 Kyushu University 5 University of Tokyo 6 National Institute of Informatics
Amazon EC2 IaaS (Infrastructure as a Service) HPCI HPCI ( VM) VM VM HPCI VM OS VM HPCI HPC HPCI RENKEI-PoP 2 HPCI 3 4 5 6 7 2. HPCI 1 HPCI HPCI HPC CS() OS HPCI 1 Table 1 HPCI HPCI Usage Scenario 1 1 (1)(2) Web HPCI Super Scheduler Resource Broker Amazon EC2 IaaS HPCI
3. OS ISV(Independent Software Vendor) Web Web Internet Fig. 1 1 Advanced software development/operation environment API HPCI HPCI HPCI ID ID Shibboleth GSI HPC HPC HPCI 4. 1 API HPCI
HPCI 24 root 4.1 VM VM VM OS 2 VM OS 3 VM VM VM OS VM VM OS OS 2 Fig. 2 Distributed environment hosting service HPC VMOS VM VM 1 VM VM VM OS VM (1) VM (2) VM (3) OS (4) OS VM VM (5) VM OS (6) VM VM
5. 24 HPCI RENKEI RENKEI-VPE (RENKEI: REsources linkage for E-scIence, VPE: Virtual Private Environment) 5.1 RENKEI-VPE RENKEI-VPE VM VM RENKEI-VPE VM VM 3 RENKEI-VPE 1 1 1 OS GSI SSH Unix RENKEI-VPE RENKEI-VPE xmlrpc RENKEI- VPE VM OpenNebula 1) OS Gfarm 2) RENKEI-VPE OS VM VM VM VM suspend/resumevm OS IP libvirt kvm VM VM 3 RENKEI-VPE Fig. 3 Architecture of RENKEI-VPE (ssh ) ssh Gfarm client VM Gfarm OS VM Gfarm VM OS Gfarm OS 5.2 RENKEI-PoP RENKEI-VPE RENKEI-PoP (Point of Presence) 3) RENKEI-PoP 1 SINET 10Gbps VPN RENKEI-PoP VM TSUBAME NAREGI RENKEI-PoP 2011 6 9 11 Gfarm v2.3.2 200TB 2 (
2 RENKEI-PoP RENKEI-VPE VM Table 2 Time for VM creation on RENKEI-PoPs using RENKEI-VPE RENKEI-PoP Time to Boot Time to Prepare Time from Prepared Time for to Boot gfexport 1 116 19 97 0.549 2 111 14 97 1.47 NII1 118 21 97 7.58 NII2 141 44 97 23.2 164 68 96 60.3 4 RENKEI-PoP RENKEI-VPE Fig. 4 RENKEI-VPE on RENKEI-PoPs NII) 2 1 5 RENKEI-VPE VM 4 RENKEI-PoP RENKEI-VPE Gfarm Metadata Server RENKEI-PoP RENKEI-VPE Gfarm File Server RENKEI-PoP VM OS Gfarm RENKEI-VPE VM RENKEI-PoP VM OS QCOW2 653MB CentOS5.5 1 RENKEI-PoP 1 2NII1NII2 1CPU 1024GB VM 2 Time to Boot VM VM (VM init ) Time to Prepare RENKEI-PoP VM Gfarm VM Swap Time from Prepared to Boot RENKEI-PoP VM Time to PrepareTime to Boot Time for gfexport Gfarm RENKEI-PoP OS Time to Prepare VM RENKEI-PoP VM VM gfexport OS RENKEI- PoP RENKEI-PoP VM RENKEI-PoP KByte Web OS 2 5.3 HPCI RENKEI-PoP RENKEI-VPE RENKEI-VPE Gfarm RENKEI-VPE Metadata Server File Server VM SINET4 VPN 1 VM
6. PlanetLab 4) 507 1091 PlanetLab slice VM PlanetLab Internet 1 VM InTrigger 5) 17 Intrigger ssh Gfarm VM OpenNebula RENKEI-VPE IaaS VM Eucalyptus 6) OpenStack 7) Nimbus 8) Open- Nebula VM VM IP VM 7. HPCI 3 24 HPCI RENKEI RENKEI-VPE RENKEI-VPE VM OpenNebula OS Gfarm RENKEI-PoP RENKEI-VPE VM VM OS 2 HPCI WG HPCI 1) Sotomayor, B., Montero, R.S., Llorente, I.M. and Foster, I.: Virtual Infrastructure Management in Private and Hybrid Clouds, IEEE Internet Computing, Vol.13, No.5, pp.14 22 (2009). 2) Tatebe, O., Hiraga, K. and Soda, N.: Gfarm Grid File System, New Generation Computing, Vol.28, No.3, pp.1 6 (2010). 3) e- POP(Point-of-Presence) // (SWoPP2010) (2010). 4) Chun, B., Culler, D., Roscoe, T., Bavier, A., Peterson, L., Wawrzoniak, M. and Bowman, M.: PlanetLab: an overlay testbed for broad-coverage services, ACM SIG- COMM Computer Communication Review, Vol.33, No.3, pp.3 12 (2003). 5) InTrigger : 2007-HPC-111pp.237 242 (2007). 6) Nurmi, D., Wolski, R., Grzegorczyk, C., Obertelli, G., Soman, S., Youseff, L. and Zagorodnov, D.: The Eucalyptus Open-source Cloud-computing System, 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, pp.124 131 (2009). 7) OpenStack Project: OpenStack Open Source Cloud Computing Software, http: //www.openstack.org/ (2011). 8) Nimbus Project: Globus Nimbus Homepage, http://www.nimbusproject.org/ (2010).