[[PageOutline]] = Aggregate Developers' Topics = At this developers session, we will discuss several issues relating to developing GENI aggregates and standardizing aggregate behavior. Specific topics are expected to include the status of the Aggregate Manager API and RSpecs, introducing the problem of a Uniform Experimenter Environment, VLAN stitching implementation status, and providing common secure authorization for tools. This session will begin conversations that will be continued at the [wiki:GEC16Agenda/CodingSprint Coding Sprint] session. == Schedule == Tuesday, 3:30pm - 5pm == Session Leaders == Aaron Helsinger, Tom Mitchell. GENI Project Office. == Agenda / Details == * Status Updates -- Aaron Helsinger, GPO * Aggregate Manager API * RSpecs * Omni * Uniform Experimenter Environment * Coding Sprint * Stitching -- Tom Lehman, MAX * Progress Report * Implementation Issues * Discussion * Secure Tool Authorization -- Rob Ricci, University of Utah == Summary == At the Aggregate Developers' Topics session, we reviewed various ongoing activities to develop and enhance aggregates, and began a discussion for some necessary new functionality. Aaron Helsinger reviewed the status of various issues: - Aggregate Manager API version 3 runs at !PlanetLab, is mostly done at ProtoGENI, and will be implemented elsewhere - Version 4 is not yet finalized, but will include Update() as previously adopted - GENI RSpecs could be refactored in future, but extensions are powerful - Omni 2.2.1 is out, and 2.3 is coming soon with stitching support - Aggregates should highlight different features, but minimize unnecessary differences - The Uniform Experimenter Environment session Thursday afternoon extended on this Tom Lehman discussed the status of GENI dynamic VLAN based cross aggregate network stitching: - Prototype deployment working now (demonstration at demo night) - Architecture exists and shows promise for future dynamic networks - Stitching Service now covers key pieces of the architecture - Omni client v2.3 will drive this process in a seamless way for experimenters, and will be released soon - There are a number of scaling and future issues to discuss, including how intermediate networks want to manager their VLANs Rob Ricci introduced the topic of making secure authorization easy for tools: - Hosted tools now need to "speak as" the user, pretending to be the experimenter - This requires experimenters to give up their private key, which is not secure - A new "Speaks For" credential and option in method calls would allow experimenters to authorize tools securely - A working group should explore the details and make a concrete proposal - There was some dicusssion of how this credential should allow the experimenter to scope the authorization - by slice, by aggregate, by operation, etc - Nick Bastin argued that the "Speaks For" option in method calls might not be needed, allowing existing v1 aggregates to support "Speaks For". == Pre-Requisites == None == Background Reading == * AM API * [wiki:GAPI_AM_API_V3 AM API v3 Specification] * [wiki:GAPI_AM_API_DRAFT Adopted and Draft AM API v4 Changes] * RSpecs * [wiki:GeniRspec Design Notes] * [attachment:wiki:GEC11RSpec:ComputeOntology-071411.pdf Compute Ontology] * [http://trac.gpolab.bbn.com/gcf/ Omni] * [wiki:GeniNetworkStitching Network Stitching Design Notes]