9/17/2013

• Shrikanth presented a comparison across 3 prominent OpenFlow controllers: ​OpenDaylight, Floodlight, and ​RYU
  • Only Floodlight doesn't support OF version 1.3 (ipv6, MPLS, multiple flow tables). Other differences minor. Programming effort not very different(?) Since RYU is python and other 2 are Java-based, migrating current MF controller code (Java) is simpler mirgating to OpenDaylight.
  • Shrikanth will clarify if MF header matching is better served any better by v1.3
• Sugang presented a summary of MF design for M2M application domain and implementation for sensor data dissemination to subscribers using mobile phone gateways
  • Outstanding items include:
• Yi presented evaluation of locality-aware scheme enhancements to GNRS
  • goal: establish a geo model to map routers/PoPs/networks, requests, mobility events to specific geo-location
  • will use a path-inflation parameter to determine routing latencies between geo-locations
  • Rocketfuel data used to establish AS-to-geolocation, inter-AS latencies (or inter-pop), and intra AS distances
  • Presented GNRS query and update models extracted from ​Alexa popularity index data and vehicular mobility traces
• Feixiong presented MF content approach using GNRS with some focus on edge-caching
  • Ray points out exploring mobile edge caching - small-cell caching - may be interesting
  • CDN requests data will be useful for evaluation - Ray might check with Akamai

9/24/2013

• Yi presented the city/PoP based network topology
  • Ray suggested the geo-distance based network topology model would be a useful full length talk
  • Could also get Jim Kurose's input on the model
  • Next steps include modifying GNRS simulation for locality and adopting the above PoP model
• Chao presented slides summarizing EIR - maybe from reading the paper?
  • Ray: look at messaging protocol incl. message fields in implementation for correctness and completeness
  • Ray: look to implement active fine-grain (per-client) protocol to query link/path quality at the edge to decide among multiple paths
  • Need to address reviewer doubts whether inter-domain is the right place to address issues at the edge
• Francesco presented an update on the host stack implementation and demo plans for GEC-18
  • Stack and API code restructured for cleaner layering - e.g., pluggable alternate transport
  • API reflects basic ops defined in MobiArch paper - missing 'get', content and service ops
  • Demo at NYU poly will be a breadcrumb or geo-tagged message drop app
    • Ray/Akash suggest that the use of QR codes for geo-tagging (location to GUID) could be confusing - need further discussion on appropriate fine-grain geo-tagging