Distributed mobility management and application discovery
- 1. Large scale RINA Experimentation on FIRE + ARCFIRE Final review Experiment 5 Eduard Grasa, i2CAT September 2019
- 2. Problem statement: mobility Large-scale RINA Experimentation on FIRE+ 2 Subnet 1 Subnet 2 Subnet 3 Subnet 1 Subnet 2 Subnet 3 Subnet 3 Subnet 1 Subnet 2 β’ App names that are stable β’ Network @ that reflect location (change) App A App B Net @ PoA Net @ PoA Mobile Host (UE) Server Access Points Access Points Access Points
- 3. Problem statement: mobility Large-scale RINA Experimentation on FIRE+ 3 Subnet 1 Subnet 2 Subnet 3 Subnet 1 Subnet 2 Subnet 3 Subnet 3 Subnet 1 Subnet 2 β’ App names that are stable β’ Network @ that reflect location (change) App A App B Net @ PoA Net @ PoA Mobile Host (UE) Server Access Points Access Points Access Points
- 4. Subnet 1 Subnet 2 Subnet 3 Subnet 1 Subnet 2 Subnet 3 Subnet 3 Subnet 1 Subnet 2 App A App B Net @ PoA Net @ PoA Mobile Host (UE) Server Access Points Access Points Access Points β’ In the Internet β App name = IP @ + transport port β Net @ = IP @ Why mobility is hard on the Internet
- 5. RINA: mobility management within a layer 1.3 1.4 1.5 1.71.6 1.1 1.2 Subnet1 (1.x) 2.3 2.4 2.5 2.7 2.6 2.1 2.1 Subnet2 (2.x) 0.1 0.2 2.81.8 5 (1) IPCP in MH @ 1.8 (2) IPCP in MH @ 1.8 @ 2.8 (3) IPCP in MH @ 2.8 1.8 2.8 IPCPs in Base Stations IPCPs in Base Stations IPCPs in edge routers IPCPs in core routers IPCPs in core routers
- 6. Mobile network with multiple layers Border Router Core DIF Under DIFs Border Router Under DIFs Border RouterBase StationMH Radio DIF Under DIFs District DIF Metro DIF Regional DIF Public Internet DIF Application-specific DIF Mobile Infrastructure NetworkCustomer Terminal β¦ β¦ β¦ Under DIFs Operator core β’ Create as many DIFs as needed (could be one) depending on density of mobile devices and speed of mobility in different parts of the mobile network β’ Each application can use the DIF that provides enough scope and QoS for its communication needs -> not restricted to the βtop onesβ β’ All layers have the same structure and protocols, implementations can be highly optimized; overhead of adding new layers is minimal 6
- 7. Experiment goals β’ Main goal: Experimentally show how the RINA structure can accommodate mobility in the simplest possible way β Without specialised protocols β Without tunnels β Without point solutions β’ Secondary goals β Explore application discovery across DIFs via the DIF Allocator β Explore support for multi-access (multiple operators, multiple physical media) Large-scale RINA Experimentation on FIRE+ 7
- 8. RESULTS: DISTRIBUTED MOBILITY MANAGEMENT Large-scale RINA Experimentation on FIRE+ 8
- 9. DMM scenario: Physical systems β’ 2 provider networks β 3 access routers (WiFi APs) β 2 Edge Routers β 1 Core router β’ An βInternetβ β 2 ISP routers β 2 servers β’ A Mobile Host Large-scale RINA experimentation on FIRE+ 9 Wifi (ssid irati) AP1 Wifi (ssid pristine) AP2 Wifi (ssid arcfire) AP3 Wifi (ssid irina) AP4 Wifi (ssid rinaisense) AP5 Wifi (ssid ocarina) AP6 LaptopMH Raspberry Pi 3B VLAN 10 VLAN 20 VLAN 30 VLAN 40 VLAN 50 VLAN 60 10 20 30 40 50 60 Edge 1 Edge 2 Edge 3 Edge 4 100 110 200 210 Core 1 Core 2 ISP 1 ISP 2 Server 1 Server 2 120 220 300 310 320 VLAN-Aware Eth. Switch Mobile net 1 Mobile net 2 Laptop running Demonstrator (Blue boxes are QEMU/KVM VMs)
- 10. DMM scenario: DIFs β’ 2 Mobile Network DIFs (1 for each provider) β’ An Internet DIF floating on top β’ Mobile Host is multi-homed (2 WiFi interfaces) β Soft handover Large-scale RINA experimentation on FIRE+ 10 MH AP 1 Edge 1 Core ISP1 Server1 Mobile network DIF Internet DIF Application Shim DIF Eth Shim DIF Eth Shim DIF Eth Shim DIF Eth Shim DIF Eth MH A1 A3 A2 E1 C1 E2 Mobile Network DIF I1 MH C1 S1 S2I2 Internet DIF C2
- 11. Managing mobility at multiple layers β’ Soft handover approach: attach to two APs briefly (from same or different provider), then transition to the new one. Exploit multi-homing. β Other handover schemes are also possible (2 WiFi interfaces not required) Large-scale RINA Experimentation on FIRE+ 11 1.1.1 1.1 MAC 1 App A ssid irati 1.1.1 MAC 1 App A MAC 2 ssid irati ssid pristine 1.1.1 MAC 2 App A ssid pristine 1.1.1 MAC 2 App A MAC 1 ssid pristine ssid arcfire 1.1.1 MAC 1 App A ssid arcfire 1.1.1 App A MAC 2 ssid irina ssid arcfire 3.4 MAC 1 1.2.1 1.2.1 3.4 MAC 2 App A ssid irina β¦1.1 1.1 1.1 1.1 1.1 Experiment time IPCPandapplication instancesatMH T1 T2 T3 T4 T5 T6 T7
- 12. DMM: impact of handover β’ Service continuity is always preserved β’ Increase in packet loss and delay due to handovers β Consequence of the prototype implementation and use of WPASupplicant Large-scale RINA Experimentation on FIRE+ 12 0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Lossprobability Samples (individual flows to server1 rina-echo- me 1-10, and server2 rina-echo- me 11-20) Loss probability, 10 parallel flow s pe r des na on app Sta c Handover 60s Handover 40s Handover 30s Handover 20s 0 10 20 30 40 50 60 70 80 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Meanofe2edelay,ms Samples (individual flows to server1 rina-echo- me 1-10, and server2 rina-echo- me 11-20) Mean of end-to-end delay, 10 parallel flows per des na on app Sta c Handover 60s Handover 40s Handover 30s Handover 20s 0 1 2 3 4 5 6 7 8 1 Goodput,Mbps Samples (individual flows) Applica on goodput Sta c Handover 60s Handover 40s Handover 30s Handover 20s
- 13. CURRENT RESULTS: APPLICATION DISCOVERY Large-scale RINA Experimentation on FIRE+ 13
- 14. Scenario: physical systems β’ Single service provider with small edge DC to host latency-critical or high- bandwidth services Large-scale RINA experimentation on FIRE+ 14 Large-scale RINA Experimentation on FIRE+ 14 UE 1 UE 2 AR 1 AR 2 AR 3 AR 4 AR 5 AR 6 ER 1 ER 2 CR 1 ISP 1 ISP 2 SRV 5 SRV 6 SRV 1 SRV 2 SRV 3 SRV 4 ToR 2 ToR 1 DC GW Small DC Service Provider net Data Center Gateway User Equipment Provider Access Router Edge Router Provider 1 Core Router ISP Router Server Top of Rack Router β’ Six access routers β WiFi Aps β’ 2 Edge routers β’ 1 Core router β’ 1 small DC β 1 DC GW β 2 TORs β 2 servers at each ToR β’ 2 Mobile Hosts (UEs) β 1 roaming, 1 static
- 15. Scenario: DIFs β’ Mobile network DIF: provides DMM and supports service DIFs β’ Internet DIF: provides access to apps available through servers at the Internet β’ Slice1 DIF: provides UE1 access to apps available through provider DC β’ Slice 2 DIF: provides UE2 access to apps available through provider DC Large-scale RINA Experimentation on FIRE+ 15 UE Access 1 Edge 1 Core 1 ISP1 Server6 Mobile network DIF Internet DIF DAF (rina-tgen or rina-echo-time) Shim DIF WiFi Shim DIF Eth Shim DIF Eth Shim DIF Eth Shim DIF Eth UE 2 A1 A4 A2 E1 C1 E2 Mobile Network DIF I1 UE C1 S1 S2I2 Internet DIF A3 A5 A6 DC UE Access 2 Edge 1 DC Gateway ToR 1 Server 1 Mobile network DIF Enterprise 1 VPN DIF DAF (Any demo app) Shim DIF WiFi Shim DIF Eth Shim DIF Eth Shim DIF Eth Shim DIF Eth S2 GW S4Enterprise 2 VPN DIF DC Fabric DIF S1 UE 1 GW S3Enterprise 1 VPN DIF UE 2 S2 S1 S3 S4 GW ToR 1 ToR 2
- 16. Application discovery (I) Large-scale RINA Experimentation on FIRE+ 16 Access 2 Edge 1 DC Gateway ToR 1 Server 1 Mobile network DIF Slice 1 DIF Shim DIF WiFi Shim DIF Eth Shim DIF Eth Shim DIF Eth Shim DIF Eth DC Fabric DIF A C UE App_name: rina-echo-time.server-aneto--
- 17. Application discovery (II) Large-scale RINA Experimentation on FIRE+ 17 UE Access 1 Edge 1 Core 1 ISP1 Server5 Mobile network DIF Internet DIF Shim DIF WiFi Shim DIF Eth Shim DIF Eth Shim DIF Eth Shim DIF Eth A B App_name: rina-echo-time.server-montblanc--
- 18. Application discovery (III) Large-scale RINA Experimentation on FIRE+ 18Large-scale RINA Experimentation on FIRE+ 20 App_name: rina-echo-time.server-ll-- Access 2 Edge 1 Shim DIF WiFi Shim DIF Eth A UE Mobile network DIF D
- 19. CURRENT RESULTS: MULTI-ACCESS SUPPORT Large-scale RINA Experimentation on FIRE+ 19
- 20. Scenario: physical systems β’ One app @ UE accessing another app at the Small DC β’ Another app @ UE accessing an app at the βInternetβ (hosted at SRV3) β’ As the UE moves, it uses different providers, DIFs and physical media Large-scale RINA Experimentation on FIRE+ 20 UE 1 AR 1 AR 2 AR 3 ER 1 ER 2 CR 1 SRV 3 SRV 1 SRV 1 SRV 2 ToR 2 DC GW Small DC Mobile provider (bzness) Data Center Gateway User Equipment Provider Access Router / CPE Edge Router Provider Core Router ISP Router Server Top of Rack Router ISP 2 ISP 3 ISP 1 CPE 1 UE 1 Fixed provider (home) CR 2 UE 1
- 21. DIFs (I) Large-scale RINA Experimentation on FIRE+ 21 UE Access 2 Core 1 DC Gateway ToR 1 Server 1 Mobile network DIF Enterprise 1 VPN DIF DAF (Any demo app) Shim DIF WiFi Shim DIF Eth Shim DIF Eth Shim DIF Eth Shim DIF Eth DC Fabric DIF S1 UE 1 GW S3Slice 1 DIF S2 S1 GW ToR1 DC DIF
- 22. UE CPE 1 Edge 2 Core 2 ISP2 Server3 Fixed network DIF Internet DIF DAF (rina-tgen or rina-echo-time) Shim DIF Eth Shim DIF Eth Shim DIF Eth Shim DIF Eth Shim DIF Eth UE2 CPE 1 E1 C1 Fixed Network DIF S3I3 I2 Internet DIF DC I1CP1 UE CP2 ISP3 Shim DIF Eth DIFs (II)
- 23. Experiment, time T0 Large-scale RINA Experimentation on FIRE+ 23 MAC 1 MAC 2 MAC 3 1 mob SSID irati 1 slice 1 inter IPC Processes at UE, T0 echo mont echo aneto rina-echo app (aneto) rina-echo app (montblanc) 2 34 5 6
- 24. Experiment, time T2 Large-scale RINA Experimentation on FIRE+ 24 MAC 1 MAC 2 MAC 3 1 mob SSID pristine 1 slice 1 inter IPC Processes at UE, T2 echo mont echo aneto rina-echo app (aneto) rina-echo app (montblanc) 7 34 5 6
- 25. Experiment, time T4 Large-scale RINA Experimentation on FIRE+ 25 MAC 1 MAC 2 MAC 3 1 mob SSID arcfire 1 slice 1 inter IPC Processes at UE, T4 echo mont echo aneto rina-echo app (aneto) rina-echo app (montblanc) 8 34 5 6
- 26. Experiment, time T6 Large-scale RINA Experimentation on FIRE+ 26 MAC 1 MAC 2 MAC 3 1 mob SSID arcfire 1 slice 1 inter IPC Processes at UE, T6 echo mont echo aneto rina-echo app (aneto) rina-echo app (montblanc) 8 4 5 6 1 fixed 10 9 VLAN 40
- 27. DEMO Large-scale RINA Experimentation on FIRE+ 27
- 28. TIP demo: systems β’ Multi-carrier slices / VPNs with mobility management support β’ Demo general view Large-scale RINA experimentation on FIRE+ 28
- 29. TIP systems: DIFs β’ Demonstrator video UE video Large-scale RINA Experimentation on FIRE+ 29
- 30. CURRENT CONCLUSIONS AND NEXT STEPS Large-scale RINA Experimentation on FIRE+ 30
- 31. No One (Tunneled) Size Fits All β’ Just increasing capacity is not the answer β Procession overheads, bloating, strategies adapted to concrete scenariosβ¦ β The risk of one-headed solutions β’ Future network services will require a blend of properties β Latency, reliability, capacity, pacing, coverage, mobility, durability β Service diversity: mMTC, URLLC, eMBB β And more to come, including on-demand β Bespoke service interfaces β On a single sustainable infrastructure β Thus the slicing requirement β’ Slices as overlays β The usual techniques to βisolateβ pieces of the network β At the price of more overhead and ossification β’ SDN and NFV may provide a more elastic way to scale β But just as a patch of the architectural deficiencies Large-scale RINA Experimentation on FIRE+ 31
- 32. Slicing RINA Style β’ Access independent services β At any edge β Inherent mobility β’ Global support for vLANs β At any layer β’ E2E management β Consistent β Low overhead β Seamless introduction of SBA and control transactions β’ Guarantee QoS/SLAs β Flow control β Bursty traffic optimization β’ Inherent Security β Implicit isolation and positive authentication β Even support for sovereign identity Large-scale RINA Experimentation on FIRE+ 32
- 33. The Sustainability Requirement β’ Feasible advanced network services β Flexibility, agility, re-use and automation β Seamless integration with management solutions β’ Software Networking β Commoditization of network equipment β’ Define the common elements in computer networking β Programmability β’ Define the variable behavior for common elements, and hence common APIs to program them β Deal with complexity β’ Maximize the invariants, hence requires far less protocols to provide computer networking β’ And, well, the network neutrality issueβ¦ Large-scale RINA Experimentation on FIRE+ 33
- 34. Results / Conclusions: DMM β’ Mobility can be supported without the need of special protocols or procedures, just using the tools the network architecture provides: β Complete naming and addressing scheme, routing updates β Designing the number and size of layers in different parts of the network to accommodate the load, scale, and rate of change of the mobile terminals β’ RINA simplifies design & operation of mobile networks, allowing greater scalability horizontally (within a layer) and vertically (through multiple layers) β’ Future work: Quantify overhead of adding more layers vs. making a single layer larger in order to accommodate a larger number of mobile hosts. Large-scale RINA Experimentation on FIRE+ 34
- 35. Application Discovery across DIFs (DIF Allocator) β’ Dynamic application discovery at multiple layers β Applications can be accessible & discoverable across multiple layers β DIF Allocator performs inter-DIF application discovery (layer directories perform this task within a layer) -> no need to replicate this functionality in application-specific protocols such as SIP β No need for user systems or apps to know what DIF they need to use (equivalent to avoiding the need to cofigure APNs in current 3G/4G networks) Large-scale RINA Experimentation on FIRE+ 35 Access Core 1 Mobile network DIF Shim DIF WiFi Shim DIF Eth DA DA UE ET s VPN DIF Before Flow Allocation Access Core 1 Mobile network DIF Shim DIF WiFi Shim DIF Eth DA DA UE ET s After Flow Allocation VPN DIF ET c
Editor's Notes
- #5: Seamless (application does not notice it) mobility is complicated due to incomplete naming & addressing: Applications need an identifier that is stable when their host moves across networks To make routing scale the network addresses need to change as the host attaches to different networks But in the Internet (layer) there is only one identifier: the IP address Special protocols to try to make it work: Mobile IP(v4/v6), Proxy Mobile IP (v4/v6), GTP for cellular (create a huge layer 2 subnet), LISP Most of them require tunnels (expensive to setup), all have limitations at the scale they can provide seamless mobility
- #6: * Link state routing within each serving area * Size of each DIF limited by # of mobile devices and speed of mobility (so that routing has time to converge) -> No problem, we can use multiple DIFs to structure the mobile network