Cloud Computing and Challenges

Download Cloud Computing and Challenges

Post on 22-Feb-2016

35 views

Category:

Documents

0 download

Embed Size (px)

DESCRIPTION

Cloud Computing and Challenges. C. Loomis (CNRS/LAL) Seminar (RAL) 13 October 2010. Biography. High-Energy Physicist (

TRANSCRIPT

<p>Operational Considerations From Running Grid Services on Cloud Resources</p> <p>Cloud Computing and ChallengesC. Loomis (CNRS/LAL)Seminar (RAL)13 October 2010StratusLab is co-funded by theEuropean Communitys SeventhFramework Programme (Capacities)Grant Agreement INFSO-RI-261552#1BiographyHigh-Energy Physicist (1 month without problems</p> <p>#User and Administrator SurveysSurveys for Users and AdministratorsCollect use cases and requirements from target communities248 and 608 unique views, 22 and 56 completedAnalysis: D2.1 (http://stratuslab.eu/doku.php?id=deliverables)User Feedback33% of users are regular cloud usersMost respondents from CS/eng., bioinformatics, and physicsMost planned for real use within 2 yearsAdministrator Feedback68% of administrators will deploy virtualization/cloud technologiesHalf have already deployed; remainder within 1 yearReluctance for administrators to trust user-generated VMs#Availability2-3 November 2010ScopeBasic cloud distributionGenerate feedback from users and administratorsWarningAll APIs, commands, etc. are subject to change!DocumentationUser tutorialDescription of reference deploymentInstallation guideServicesAppliance repositoryUser and administrator supportSoftwareRepackaged/patched OpenNebulaCommand line tools (user and administrator)Standard images (ttylinux 9.4, Ubuntu 10.04, CentOS 5.5)Web monitorManual and Quattor-based installationInitial Public Release#Appliance RepositoryApplianceVirtual machine created to provide a single service or a few tightly coupled servicesAppliances can be used to exchange software and knowledge Appliance repositoryRepository allows virtual machines to be found and sharedCurrent implementation just a simple web serverContains base images, will have grid images as wellFuture directionsWorthwhile to use metadata schema like Dublin Core?Publish metadata via RDF for others to use?Federation of separate appliance repositories?</p> <p>#Site Management ToolsQuattorUse to demonstrate integration with site management toolsChosen because of partner experience and use within EGINearly complete automated installation with Quattor available</p> <p>Virtual machine metadataInformation about virtual machine contents critical (esp. when shared)Tools like Quattor can be used to document those virtual machinesCan be used as mechanism for generating images as well</p> <p>#Meeting ExpectationsBare Metal PerformanceVirtualization tax near zero from technological improvementsIO still issue, but also on physical resources as well!Instantaneous (~10s) Availability of MachinesFast copies, caching of machine imagesClient site preparation of contextualization image, Optimization of OS boot sequenceFully Elastic CloudsUsers expect infinite resources, machine room reality is different!How to prevent users from seeing a queue?Allow resource searching/matchmaking like the grid?Site-level, automated bartering through hybrid clouds?#Sharing ResourcesPossible to bring grids collaborative spirit into the fundamentally selfish cloud paradigm?</p> <p>Authentication/Authorization MechanismsReuse worldwide trust network and its technologiesMachine Images and AppliancesGreat way to share algorithms, software, packaged data, Issues of transport, caching, and access controlData ManagementLike grid, issues of access, persistency, versioning, Additional issue of service state, logging, accounting, #Building (More) TrustReluctance to Run User-Defined ImagesRealizing full potential of cloud infrastructures will require building more trust between users, VOs, and system administratorsTrust will vary with actors involved sliding scale that balances level of trust with allowed capabilities</p> <p>RequirementsDefine security requirements for images (HEPiX)Matchmaking: appliance metadata vs. site requirementsEnhanced monitoring (e.g. ports and connections)Additional trust between firewall controllers and site administrators#Impact in Scientific CommunitiesNo shoehorn needed!No need to port application to specific framework or to create adaptors for different environments Sharing without requiring specific OSes, software, versions, Specialized platforms AND efficient sharingPlatforms exist now with static deployments and allocationsReconfigurations impede sharing of underlying resourcesCloud makes process more agile and dynamicGreater accessibilitySmaller groups can outsource IT deployment/maintenance Modest calculations done quickly: no procurement, installation</p> <p>Greater responsibility for IT engineering &amp; development#ConclusionsGrid and cloud technologies are complementaryEasier deployment, maintenance of sites for administratorsMore flexible, capable, and dynamic infrastructure for usersStratusLabEarly results: surveys, functioning grid services over cloud, appliance repositoryRelease: expected in a couple of weeks, accessible cloud infrastructures for testsChallenges for cloud communityMeeting expectations of users and administratorsSharing easily the diversity of resources in communitiesBuilding more trust between actors#Copyright 2010, Members of the StratusLab collaboration: Centre National de la Recherche Scientifique, Universidad Complutense de Madrid, Greek Research and Technology Network S.A., SixSq Srl, Telefnica Investigacin y Desarrollo SA, and The Provost Fellows and Scholars of the College of the Holy and Undivided Trinity of Queen Elizabeth Near Dublin.</p> <p>This work is licensed under the Creative CommonsAttribution 3.0 Unported Licensehttp://creativecommons.org/licenses/by/3.0/#</p>