| 34 | ==== Meeting Summary ==== |
| 35 | '''Tom Lehman''' reviewed the agreed-upon GENI stitching architecture: |
| 36 | - Several functions, to be implemented in various locations or |
| 37 | skipped in certain scenarios : |
| 38 | - Static topology collection |
| 39 | - Path computation |
| 40 | - Workflow/coordination |
| 41 | - A common schema for describing connections between aggregates, |
| 42 | and the capabilities of those connections. |
| 43 | - AM API revisions to support 'chain' (AM to AM) mode, and |
| 44 | negotiation. |
| 45 | |
| 46 | '''The stitching schema''' |
| 47 | - Based on and interoperable with IDC / OSCARS / perfSonar Topology service |
| 48 | - AMs advertise connection points |
| 49 | - Requests and manifests list hop by hop paths |
| 50 | - http://hpn.east.isi.edu/rspec/ext/stitch/0.1/stitch-schema.xsd |
| 51 | - Follow links in Background Reading for examples and more |
| 52 | information |
| 53 | |
| 54 | '''Aaron Helsinger''' and '''Chris Benninger''' did a demo of a |
| 55 | tree-mode (client-driven) VLAN stitching from ProtoGENI (using the |
| 56 | schema) across ION, to MAX. The script calculates dependencies and |
| 57 | orders its calls appropriately. |
| 58 | |
| 59 | '''Tom''' briefly discussed some remaining points to cover |
| 60 | - Negotiation: How do 2 Aggregates negotiation a common stitching |
| 61 | mechanism and configuration? IE, how do they pick a VLAN tag? |
| 62 | - Proposal: an AM API call that reserves a resource without |
| 63 | instantiating it, returning the reserved resource and the other |
| 64 | options that are available |
| 65 | - Chain mode: need an aggregate to aggregate call |
| 66 | - Note authorization issues |
| 67 | |
| 68 | ===== Comments and discussion points ===== |
| 69 | - Not all architecture components are required. |
| 70 | - AMs should be able to accept the complete request RSpec at each AM, |
| 71 | allowing them to use extra information or ignore it. |
| 72 | - The schema cannot refer to existing RFCs directly, due to some GENI |
| 73 | customization (like identifier format). |
| 74 | - GRE tunnels and other connection technologies must follow. |
| 75 | - !OpenFlow support is not there - yet. |
| 76 | - Multipoint paths are not supported in the underlying services |
| 77 | yet. When DCN/OSCARS supports that, we can. But with a bunch of |
| 78 | pairwise paths you can still create complex topologies. |
| 79 | - ProtoGENI implementation of the schema is incomplete. |
| 80 | - Rob Ricci expressed an interest in a production ION aggregate to |
| 81 | encourage experiments to use his ION connections. |
| 82 | - There was vigorous discussion about the feasibility and |
| 83 | difficulties in having such a real ION aggregate (political, policy, |
| 84 | time, money, more than technical). |
| 85 | - How does ION track who is reserving the circuit? |
| 86 | - Will policy allow any GENI experimenter to reserve any circuits? |
| 87 | - Are there sufficient connections through regional networks to make |
| 88 | this possible in general? |
| 89 | - This discussion of stitching does not cover: |
| 90 | - node-to-the-first-switch or other connections that are technology specific or internal to an AM, |
| 91 | - multi layer dependencies, or |
| 92 | - application layer dependencies (client/server). |
| 93 | - Orca has a notion of dependencies that does cover these more general cases. |
| 94 | |
| 95 | ==== Possible Next steps ==== |
| 96 | - Make sample stitching client code available |
| 97 | - Complete schema implementation at ProtoGENI |
| 98 | - Discuss and agree on APIs to support negotiation (eg of VLAN tags) |
| 99 | - Investigate policy considerations in having a real ION Aggregate |
| 100 | Manager |
| 101 | - Prototype topology service based on perfSonar topology service |
| 102 | - Prototype path computation service based on IDC computation service |
| 103 | - Discuss and agree on APIs to support chain/AM-to-AM mode |
| 104 | |