[lisp] IETF115 LISP-NEXAGON deck & notes

[Sharon Barkai <[e-mail]>]

IETF115 LISP-NEXAGON Update


Updates to lisp-nexagon since Vienna. 2 updates relate to AD review and 2 to relate to industry developments associated with the draft
Next, The draft combines geospatial grid IDs and LISP EIDs for mobility geolocation networking. There where two choices for formal grid: S2 and H3. S2 is more geometry mapping atlas GIS oriented. H3 better suited for dynamic realtime conditions, realtime network, aggregation and addressing. Simple hierarchy and neighbor propagation. As requested by AD We added an LCAF for clearly encoding H3 IDs in EIDs with the kind help of Dino an Alberto
Next, the network is addressed and packets are keyed by H3 IDs for Values based on BDD taxonomy. It was requested by the AD that these two consortiums review the draft. H3org was excited about the new networking use for the hierarchical grid but pointed out one issue related to hexagonal hierarchy. BDD was asked to confirm the enumerated detections and assumed localization abilities by moving vehicles with reasonable onboard compute stated in the draft. Offline BDD added an enumeration enhancement. Both of these are now addressed by the draft.
Next, an interoperable multi party PoC was made public by the AECC using the draft to find free parking spots in Tokyo. KDDI provided access, Oracle provided edge compute locations. The network locations were tied by hosted RTRs. Nexar provided both the AI car-cams and AI consolidation but this is unlikely to be the case, multi MSPs multi gear vendors more likely. RTR PCAPs where used to prove: Steering uploads to right nexagon location per H3 EIDs detection location, subscription to Nexagons per H3 EIDs areas of interest, subscription continuity and geo-privacy per Vehicle EIDs. The PoC is now extended to include more detections, vehicles, and geospatial areas.
Next, the next public PoCs will dramatically increase the Fog fragmentation axis of the network testing. Cisco breakouts will be used to upload and subscribe to high distribution of edge locations in Toyota gas-stations and car dealerships. This will push the LISP aspects much further and reflects industry trends for: Green, Low cost, High capacity connected mobility.
Next, additional industry related connected vehicles trend is Automotive Digital Twins in general ( in addition to nexagon street segments twins). Digital Twins in general are based on connected sensors to compute objects reflecting physical assets. Not much related to LISP, but if we examine 5 points in the slide regarding DTs in closed domains like a factory vs an open field like a city we see clearly the LISP/CAN applicability. To those we add the Fog fragmentation which is critical for Reality+ economical viability.
Next, If we look at the Fog fragmentation supported by draft-lisp-nexagon we see that EID geospatial twins can extend across edge GPU locations per street load - which are priced much lower, 35 cents per kWh , than cloud GPUs (no spine, H2 power/cooling, no monopoly). We also see that vehicle EIDs can freely roam between breakouts and do not rely on monolithic carrier spectrums. Breakouts are priced per UE and not per GB like carrier IOT. Combined these are important for Reality+ economical viability.
Next, If we examine the generalized structure of using lisp-nexagon for a network connecting multi OEM vehicles to multi vendor twins we can see the difference in economics, for 1k FPS, resulting in a network and compute bill, which is reduced from $1.25 Ms 15x to less than $100k per month
Next, this slide provides detailed benchmark of 1k FPS per city. ADAS use cases will increase these economic diffs by 10x factor



--szb
Cell: +972.53.2470068
WhatsApp: [phone]