Middleware for Grid Computing On Virtual Machines Renato

Middleware for Grid Computing On Virtual Machines Renato Figueiredo José A. B. Fortes Ivan Krsul, Sumalatha Adabala, Vineet Chadha, Andrea Matsunaga, Ziad Saleh, Mauricio Tsugawa, Jian Zhang, Liping Zhu, Xiaomin Zhu Peter A. Dinda Ananth Sundararaj, Ashish Gupta Advanced Computing and Information Systems (ACIS) Elec. and Comp. Eng. , University of Florida Prescience Lab Computing “In-VIGO” (In Virtual Grid Organizations) Comp. Sci. , Northwestern University Middleware challenges n n Virtual Machine = process (VM monitor) + data (VM image and user files) Image and Data Management n n Resource Management n n Creation, storage, and transfer of VM images for dynamic instantiation, and user data Application and resource perspectives Virtual Networks Data management: Grid virtual file systems n Virtualization: Virtual resources Polymorphism Manifolding Multiplexing n n Physical resources “Classic” Virtual Machines (VMs) n Many, distinct O/Ss that multiplex a physical resource E. g. VMware (x 86), IBM z/VM (S/390) n Grid computing with n n Resource management n n Relational database queries Resource control via real-time schedules Prototype and applications n VMs 1 Security, isolation Customization, legacy support Resource control, site independence n n Seamless access to decentralized storage services of a grid User-transparent virtual file systems created on-demand on top of NFS (V 2, 3) Supports unmodified binary applications and native O/S clients, servers n n In. VIGO middleware beta (Spring 2003) Java, Globus, SQL, Apache, VFS Netcare, n. Cn, Digital Government, BMI, LSS Wide-area Testbed Architecture ‘A’ Virtual back-ends ‘B’ ‘C’ Users: n. Cn, Netcare, … A Service provider ‘S’ V 1 B V 2 C V 3 Y V 4 X Internet 64 -processor IBM x. Series, VMware Data Server D User ‘X’ Information service 1 Front end ‘F’ P 1 Internet (Abilene) P 2 64 -processor IBM x. Series, VMware Physical server P IBM z 800 middleware VM startup data session Image Server I “A Case for Grid Computing on Virtual Machines”, R. Figueiredo, P. Dinda, J. Fortes, Proceedings of ICDCS 2003 3. 4 TB IBM “Shark” University of Florida Northwestern University Sponsors Research grants by the National Science Foundation: ANIR NSF Middleware Initiative (NMI), EIA CISE-RR, and by IBM