Virtual Infrastructure in the Grid Kate Keahey Argonne National Laboratory
01/30/06MSI training Event The Grid Metaphor
01/30/06MSI training Event The Grid Metaphor
01/30/06MSI training Event The Grid Metaphor How do we store energy? How do we charge for energy? How do we reliably deliver energy? What happens if a power station fails? How do we ensure quality of service? What elements make for a safe and efficient power Grid? How do we make sure that supply meets demand?
01/30/06MSI training Event Providers and Consumers l Providers u Own, operate, and contribute physical resources u Require incentives to participate u Low participation costs u Protection from activities of the consumer u Ability to control and monitor resource usage l Consumers u Want on-demand access to computational resources at modest cost u The ability to configure them to meet their needs u Reasonable guarantees of resource availability l Scalability u Provider and consumer roles have to be decoupled
01/30/06MSI training Event Requirements for a Grid Execution Environment l Environment and configuration u A VO should be able to provide the configuration it needs independently of the resource provider l Isolation u The provider needs to be able to delegate resource usage to the VO so that the VO can’t impact the resource provider -- and thus does not need to be under its control l Resource usage and accounting u The provider needs to be able to grant, enforce and account for VO resource usage in a way that is independent of how the resource is consumed l All of this must be available on-demand!
01/30/06MSI training Event Virtual Workspaces A virtual workspace is an abstraction of an execution environment that can be made dynamically available to authorized clients by using well- defined protocols. l Two dimensions: u Software configuration u Resource quota (CPU, memory, etc.) l Examples of Workspaces: u A physical machine configured to meet TeraGrid requirements u A cluster of virtual machines configured to meet OSG requirements u A cluster of physical machines running a hypervisor
01/30/06MSI training Event Virtual Workspace Implementations l Physical resources u Allocate and configure a physical resource l Cluster on Demand (COD), Duke University l Bcfg project at ANL u As a method they are inflexible and coarse-grained l Virtual resources u Allocate resources for and deploy configured virtual machines l Existing efforts: In-Vigo, Virtuoso, VIOLIN, the workspace project… u Much more flexible, allowing migration and fine-grain enforcement.
01/30/06MSI training Event Virtual Machine Basics Hardware Virtual Machine Monitor (VMM) / Hypervisor Guest OS (Linux) Guest OS (NetBSD) Guest OS (Windows) l A VM can serialize all of its state (including RAM) u A VM image is simply a collection of files l Disk partitions, RAM, configuration file u Such image can be easily moved (migrated) between hypervisors of the same type u Such image can also be saved and used for rollbacks VM App
01/30/06MSI training Event The Need for Speed LXVU SPEC INT2000 (score) LXVU Linux build time (s) LXVU OSDB-OLTP (tup/s) LXVU SPEC WEB99 (score) Benchmark suite running on Linux (L), Xen (X), VMware Workstation (V), and UML (U) Paper: “Xen and the Art of Virtualization”, SOSP 2003
01/30/06MSI training Event Summary: What Makes VMs a Great Workspace Implementation l VM properties: u Excellent isolation l Generally enhanced security, audit forensics u Fine-grain enforcement potential l Details depend on implementation u Customizable software configuration l Library signature, OS, maybe even 64/32-bit architectures u Serialization property l VM images (include RAM), can be copied u The ability to pause and resume computations l Allow migration l How do we make VMs available over the network and manage them so as to leverage this potential? u Challenges: security, enforcement, protocols
01/30/06MSI training Event Deploying Workspaces in the Grid Workspace Wizard (VW Factory) Workspace Management Service (VW Repository) Workspace Service (VW Manager) request a workspace workspace meta-data manage workspace environment workspace metadata terminate workspace deployment negotiate workspace deployment manage/monitor/renegotiate workspace deployment manage activities within the workspace Workspace
01/30/06MSI training Event Workspace Implementation l Protocols: Web Service Resource Framework (WSRF) u An extension of Web Services u Standard mechanisms for creation, inspection, notification, lifetime management u Globus Toolkit 4 implementation l Provides secure authentication, authorization as well as tools for fast transfer, replica management, monitoring, and others. l Creating a workspace u workspace meta-data (workspace image) u deployment descriptor (resource allocation) l Managing a workspace u renegotiate resource allocation u Standard WSRF management functions l Challenges: resource assignment, negotiation, etc. l To download visit
01/30/06MSI training Event Putting it All Together RRRRRRRRR ACB VM1VM2VM3VM4 resources deployment capabilities virtual machines jobs B DBE deployment capabilities l Deploying a workspace requires and creates a deployment capability u Required capability is described in workspace pre- requisites l Workspaces can be layered
01/30/06MSI training Event Applications: Edge Services (1) l Edge Service: service executing on the edge of private and public network l ESF Requirements u Diverse configurations, easy to upgrade u Good potential for managing resource allocation l Status: u Testbed: SDSC, FNAL, UC u Multiple base images have been developed u One Edge Service deployed u Workspace Service developed l Timeframe: ~few months
01/30/06MSI training Event Applications: Edge Services (2)
01/30/06MSI training Event Applications: Virtual Clusters l Extends the abstraction of a workspace to a virtual cluster l Deploys a cluster on a site u SLURM, PBS/Torque implementations u Configures networking, shared storage u Image propagation, main deployment cost l Tech report available
01/30/06MSI training Event Virtual Cluster: OSG Applications GADUfMRI MontageFOAM
01/30/06MSI training Event Conclusions l In order to grow, we need to scale l In order to scale, we need to provide a reliable tool for separating producers and consumers l Virtualization provides a useful, scalable tool to decouple providers and consumers u Workspaces as physical resources u Workspaces as virtual machines l Looking forward u Grid economies