Progress Report for the GENI Experiment Engine: GEC 19
Contract Deliverables
- Submitted design document to GPO on GEE Filesystem Design
- Submitted design document to GPO on GEE Network Design
- Submitted design document to GPO on GEE Message System Design
- Demonstrated GENI Experiment Engine Slicelet Creation, experimental run, and teardown in under five minutes at GEC-19
Technical Progress
- Preallocated four-node InstaGENI PlanetLab slicelets as GENI Experiment Engine Slicelets
- Developed demonstration-quality webserver to allocate and reclaim GENI Experiment Engine Slicelets, using Google-based authentication
- Integrated Swift storage engine with Syndicate meta-data server
- Developed architecture for creation of use-once ("burner") slice credentials and delivery of credentials to user
- Integrated Fabric as deployment and experiment execution engine
- Established a Github repository for GEE code https://github.com/rickmcgeer/geni-expt-engine
Ongoing Issues
- We are highly reliant on GENI VLANs for the networking elements of the GENI Experiment Engine, and we need a guarantee that the VLANs that we allocate will be stable over, essentially, the lifetime of GENI -- at a minimum, we require some notice before a VLAN is torn down
- Need to continually renew the GEE slice on GENI, and reallocate slivers. The GEE, particularly the GEE Compute Engine, is designed as a set of long-running, stable services, not short-run experiments.
Outreach Activities
- Rick McGeer "Distributed Clouds" keynote presentation at workshop in Dresden
- Rick McGeer "Distributed Clouds" seminars at TU-Berlin and TU-Kaiserslautern
- Rick McGeer "Distributed Clouds" lecture at San Jose State University
- GENI Experiment Engine presentation at GEC-19
Plans for GEC 20
- Bring filesystem and message system to demo-quality implementation
- Expand range of actions offered through the website -- show user slicelet status, permit recovery of credential zip file
- Remove a number of security holes in website -- user email passed in plain-text parameter, in particular.
- Add hook to permit user to automatically report a bug to GEE team
- Recruit a small number of experimenters to participate in GEE beta program
- Investigate design alternatives for GEE Slicelet hosts. We currently run in PlanetLab nodes, but could potentially do this in OpenVZ ProtoGENI images, and/or OpenCloud images
- Investigate sharing raw PC with VTS. Currently, we need one raw PC per rack, and so does Nick for VTS. That's 40% of the capacity of an InstaGENI rack, which is an excessive use of resources. In both of our cases, we don't need a full PC; we just can't live in a VM.
Last modified 10 years ago
Last modified on 03/27/14 02:20:34