Cluster D Review July 2nd, 2009 GPO: Chip, Harry, Heidi, Aaron PIs: Prashant Shenoy (vise), Michael Zink, David Irwin, Ilia, Brian Lynn, Rajiv Rambanth?, Brian Irvine, phone: Karen Bergmen, Hongwei, Franz (Columbia), Caroline (Columbia), Michael (Columbia) Slides, agenda on wiki at http://groups.geni.net/geni/wiki/ClusterDMtg070209 ====== Harry: Introduction (from slides) - review of spiral 1 goals - for GENI/ORCA clearinghouse - for each substrate - GENI goals for spiral 2 - lots of live experiments - continuous operation (may be challenging) - identity management - improved integration of data and control planes - instrumentation - interoperability of clusters, permitting clusters to access the widest number of aggregates - Process - aim for GENI spiral 2 goals - start with ORCA integration roadmap - add input from all projects & GPO - to be completed by mid-July - want consensus of GPO and all PIs - will drive spiral 2 milestones for each project ====== ORCA/BEN, Ilia (from slides) Progress so far - deployed ORCA into BEN - Demonstrated VLANs across BEN using ORCA at GEC4 - Developed drivers for BEN - Developed NDL-OWL RSpecs - Connected BEN to NLR FrameNet via RENCI and Duke - Collaborated with Cluster D: software releases, ORCA-fest, integration assistance Challenges - Resource representation for heterogeneous substrates - Stitching of slivers to slices; requires information not available until after provisioning, true for networks and VMs - Connectivity issues: delays in dealing with campus IT, cost Integration process - Produced an integration document that outlines how projects of different types can choose to integrate - Code releases: tarball - Clearinghouse to be stood up in July after GEC5 demo Backbone connectivity - NLR connectivity sorted out at two BEN PoPs; 10 VLANs available from NLR (only 2 remapped by Duke) - Working on getting I2 DCN connectivity; talking to John Balbrecht at I2 for research; aiming for demo before 10/31; some costs from local provider, 100 miles between I2 drop-off and research triangle; I2 DCN people are different from the network engineering people and are different from the GENI wave Spiral 2 plans - Ontological resource representation; working with Keren on measurements and Cluster E&D on wireless - Generic solution to sliver stitching - Introducing measurement devices into BEN and adding ORCA support - Identity management using Shib - Cloud substrate and control - Feature roadmap items Progress against milestones - Planning to make ORCA a production capability for BEN; 2 issues: will the software work? and will there be sufficient documentation that people can use it? - Can ORCA be used to request connections to I2? Yes, working with Chris Tracy. More than 1 way to do it. Could create an ORCA AM for I2. Complicated by I2's work on their next-gen network architecture. - Experimenters in mind for spiral 1? Mostly Keren, already share cluster & interests. More detail later today. Other experimenters will need patience and expertise. Plan to take a cautious approach. - Could move faster with more funding. Could GPO help with documentation? Maybe a little but probably not. ORCA Milestone Review, Harry - on geni.net wiki page - 6/1 milestones will be done by 7/7 demo ======= DOME, Brian Levine (from slides) Progress in year 1 - hardware upgraded, OS virtualized, wifi virtualized - basic ORCA setup on buses, not integrated into clearinghouse yet - v1 software release started, completed portal, experiment control - preparing for demo later this summer - started cluster integration Integration with ORCA - developed a controller, creates lease, and handler, that performs actions based on leases - need DOME portal interfaces to ORCA - ORCA instance will be shared with VISE project on geni.cs.umass.edu Connection to Internet2 - In discussions with local OIT staff - Will connect geni.cs.umass.edu to i2 directly via fiber from campus to Springfield to Boston, working in tandem with VISE team; requesting 1 VLAN ID, expecting little traffic; will have the capability to inject frames with VLAN tags Experiments by outsiders - Have portal for job submission, one resource: wifi card, 900MHz radio coming next year; gets status; interfaces to ORCA - Experiments on Buses. DTN tolerant downloading of experiments to buses; dynamic creation of VM sandbox to execute experiments, includes partitions, networking, devices; scheduling of experiments based on ORCA leases - Instrumented measurements. Half the core is about remote management and diagnostics. Researcher data is uploaded automatically if certain file format is used. Plans for Spiral 2 - Goal is to deploy on buses this summer. 5-6 out of 40 buses in field per day. - Will bring on undergrad projects in the fall from "Intro to Networking" class. Imagine only 1 or 2 projects will get far enough to use the real system. Have a PhD graduating this year heading to Arkansas who is expected do experiments. - Plan to allow experimenters to access XTend radios - Want 'true and open access to testbed by experimenters' - Tutorials on how to use? May require some separate funding. Keren very interested in such a tutorial. Concern about maintaining documents after publishing. This is made more difficult since the prototypes are rapidly evolving. Concern that supporting users will delay system development because of the need to have a stable system. GPO thinks projects have latitude to define the level of stability and support that makes sense for them. Currently, GENI is in the 'exploratory prototyping' phase, likely that composition and design will change over time. Anish: We've found that tutorials are good as a way to train PIs grad students and colleagues. There's a real cost to keeping it up but it is in general a good idea. What events would be a good fit? Having a pull from users helps a lot. Chip: one approach is that folks want to use a remote system initially then make their own local copy. Anish: yes, that's our experience but doing 'cross-experiments' has been quite hard. Not clear what the benefit is to other users. What are the GPO's incentives to get people to do experiments? Not much. The GPO doesn't sponsor research. This is what NSF does. The GPO sponsor trial experiments to help shake down the system. Milestone Review, Harry (from wiki) ====== VISE, Preshant Shenoy with assist from David Irwin (from slides) year 1 progress - completed assembly, deployment of 3 nodes - initial ORCA integration - GEC4 demo (no radars but pan-zoom cameras) - outreach with UPRM underway - clearinghouse integration underway - sensor virtualization underway with some issues - can virtualize an actuator in a guest VM - testbed available for use within cluster Challenges: year 1 - Getting up to peed on radar/sensor interfaces; upgrading nodes to support multiple users; lost domain expert; simplifying radar/sensor control code. - Radar/sensor virtualization. problems controlling some devices in Xen and other virtualization domains; temporarily switching to VServers to continue progress; Xen folks seem to be fixing the problems. ORCA integration - Developing plug-in points: resource handlers, slice controllers, table-driven allocation policy - setup geni.cs.umass.edu with DOME; will host both DOME and ViSE actor servers; will transition from using a local clearinghouse to the one at RENCI when necessary Internet2 - Multiple meetings with UMass-Amherst OIT; working to get VLAN connection over NEREN to an I2 PoP in Boston (NOX). - What happens to our traffic at NOX? Ilia introduced ViSE to John Volbrecht at I2 DCN. Coordinating with him one I2 PoP connection is in place. Plans: Experiment Examples - Sensor-centric experiments. E.g., comparing fidelity from Furuno, Raymarine, MA1 (students from Peurto Rico) to focus on ground truth verification and long-term data collection with the UMass trace repository. Peurto Rico very interested in low-infrastructure radars due to its geography. Have a PhD student going back to PR, might be a good GEC grant candidate. Virtualization of these systems might be of great interest to engineering communities, not just networking. PR is developing a student testbed with tight collaboration with UMass. - Long-distance wireless experiments such as looking at wireless BW over long-distance. One REU and one UMass undergrad. Doing experiments now, would like to migrate to GENI. - Longer term vision. Connect sensed data to cloud-based storage and processing. Run complete experiments: sensing + storage + processing. BEN is applying for some processing; has a state project for micro-rain radar Current Spiral 2 Milestones - 12/2009 - sensor slivering - VISE integration with clearinghouse - 1/2010 - installation of rapidly-deployed node - installation of camera devices - 3day class at univ of PR on virtualization & GENI - 4/2010 - virtualize camera devices - integration of slivering into TB - TB allocation policy for sensors - experiment control framework from updated reference software - 10/2010 - demo with multiple experiments - make federated TB available to outside GENI users - Some spiral 2 milestones accelerated to spiral 1 Discussion - At a high level, looks like the VISE and DOME milestones are independent from ORCA. Is this true? No. Original milestones pre-date ORCA-awareness. Big ORCA contribution is in resource representation. ====== KanseiGeni, Anish (from slides) - 2 primary activities this year: - phase 1 GENIfication: required re-factoring component & aggregate managers as web services - phase 2 ORCAfiction: decomposed researcher portal to introduce ORCA actors, implementing some parts, specifying some policies for RSpec definition, - Experiments: tend to be either those interested in a) understanding low-power networking phenomena, e.g., security without shared keys requiring modification of MAC layer or b) those using sensors, e.g., collecting sensor data in a portable array then injecting it into a sensornet application running in the testbed, can be used for protocol evaluation or tuning. A recent focus has been on energy monitoring. - External users at UCLA, Northwestern, UT Dallas, ICT Australia, Michigan State, Wayne State, SUNY Buffalo - Neteye in use by universities in the US, Hong Kong, China and maybe Bangalore. - Motivated by federation scenarios: multi-fabric sensing, seamless regression testing, portable and stationary arrays, and fabric-via-cloud Challenges - What is the motivation for engaging external partners in GENI? - Overly-constrained budget. Interested in science education in later spirals; additional capabilities in D&P proposal; still interested in SunSPOTs. NSF doesn't have a way to pay for operations costs. Might want to have a workshop or other organized forum educate CISE on operations costs. - Effort needed for GENI-fication and ORCA-fication. Got a lot of help with ORCA-fication. Internet2 connection - Can get VLAN connect into Chicago PoP. Would cost ~$5k/mo. 100Mbps would be lower, cost unknown. Just an L3 node is $235/mo. At Wayne via MERIT L3 is available at no cost, L2 too high at $30k/yr. Spiral 2 plans - Focus is on federation. Need to make netEye KanseiGENI compatible. - Experiment interaction user service, i.e., GENIfication of Kansei researcher portal - Basic federation of resource discovery, embedding, and scheduling - Support for experiments, make Kansei ORCA integration model available to other testbeds. Milestone Review, Harry (via wiki) - split some milestones to reflect work accomplished and highlight ORCA integration necessary ================= Embedded Real-Time Measurements, Franz Fidler (from slides) - Motivation: Emphasis on monitoring PHY layer conditions; seeing interest by others in making measured data available. Networks will have greater diversity of bit-rates, waveforms, as well as dynamic optical routing and cross-layer optimization. Flexibility will be needed by PHY layers will cause relaxed performance constraints and that will create the need for greater monitoring. - GENI challenges: integration of measurement resources into substrate and control plane frameworks. - Progress in year 1: - assessed requirements for real-time measurements within future GENI infrastructure - Assessed interface specifications based on GENI requirements - proposed unified measurement framework (UMF) - performed performance simulations - joined cluster D - Working towards I2 connection via perfSONAR - Developing plans for experiments by outside GENI researchers - Year 1: unified measurement framework requirements & example - Plans for Spiral 2 - drive prototyping forward: hardware part of UMF and demonstrate interface between UMF and a network device capable for PHY layer monitoring - start integration efforts with BEN, who has fiber switches ====== Remaining tasks for Spiral 1 - see Harry's notes ====== Discussion on monitoring - Jeff: hyperic may be a good tool that uses ganglia to monitor VM status, can be extended to other things than VM; have a student working on this; looking for input as to whether this is a good direction. Ilia: have looked at PCP for a monitoring system. some concern that Hyperic may have an uncertain future as it was acquired last month. have indicated to GMOC that there is a strong preference to developing a SOAP interface Heidi: need to think about what data you want to expose and share Anish: Kansei has an interface for getting experiment and substrate status. ======= Discussion of Spiral 2 Capabilities (see Harry's notes) - Chip: central goal of Spiral 2 is live experimentation; won't be easy; everything else is secondary - Experiments: Want to see experiments of end-to-end systems, e.g., matching networks with servers, sensors with clouds, content with buses, radars & buses - Identity management: DOME plans to plug into some foreign authentication manager. GPO plans to try to go to Shib & InCommon, since it looks like it might take hold broadly. Rough mental model is that in the early days, almost anybody can use this. Need ways to say 'these people are out' but probably don't need elaborate policies early on. GPO is interested in knowing whether folks believe moving to Shibboleth is a bad idea. Shib has been most successful for managing user access to a portal, seems good for GENI as far as that goes and that's what ORCA uses it for. What that user is allowed to do is another issue and not currently handled by Shib attributes. Shib is only starting to handle delegated authority, still early days. - Improved integration. Need to improve distribution of keys to containers. Currently manual, ugly, error prone, requires documentation but will work until container reaches 10's. Currently no privacy. Could be added by using HTTPS on transport. - Measurements DOME collects location information on buses but needs a solution on providing it to GENI users. Has also been thinking about longitudinal studies, Brian will send Chip a copy of a paper. ERTM will prototype hardware, interface it with BEN, integrate with ORCA. - Interoperability. Want to avoid CF balkanization.