Richard - IFIP Networking 2021 - Panel.pdf
0 likes•15 views
Richard Li discusses limitations of IPv4 and IPv6 for 5G, B5G, and 6G mobile network applications. He notes that IPv4/IPv6 yields huge bandwidth waste for mMTC, UCBC, HCS, and short texts due to large packet overhead. Additionally, IPv4/IPv6 cannot guarantee key performance indicators like latency and packet loss required for uRLLC and RTBC. As an alternative, Li proposes an incremental evolution of IPv4/IPv6 that includes flexible addressing systems, geography-based addressing, and integration of satellite and terrestrial networks to expand its applicability for future applications and services.
Richard - IFIP Networking 2021 - Panel.pdf
- 1. Futurewei Technologies, Inc. Limitations of IPv4/IPv6 for 5G/B5G/6G Mobile Network Applications Richard Li Chief Scientist, Network Technologies, Futurewei, USA IFIP Networking 2021 Panel: BEYOND 5G SERVICES: PRESENT TECHNOLOGICAL LIMITATIONS & FUTURE OPPORTUNITIES Wednesday, June 23, 2021
- 2. 2 Does IPv4/IPv6 have any limitation for 5G/B5G/6G applications and services? eMBB mMTC uRLLC UCBC RTBC HCS eMBB mMTC uRLLC 6G 2030s 4G 2010s 2020s 2025 IPv4 IPv6 1981 1995 5G B5G ? ? ? ? ? ? Do we have any other choice?
- 3. Page 3 IPv4/v6 has been used as a solution for mobile network applications App (user) TCP (user) IP (user) PDCP RLC MAC PHY App (user) TCP (user) IP (user) PDCP RLC MAC PHY App (user) TCP (user) IP (user) GTP-U (S1) UDP (Nwk) IP (Nwk) IP/MPLS Backhaul Eth/Nwk App (user) TCP (user) IP (user) GTP-U (S1) UDP (Nwk) IP (Nwk) IP/MPLS Backhaul Eth/Nwk App (user) TCP (user) IP (user) Radio Access Network Fixed IP Backhaul Network IP/MPLS Backhaul Eth/Nwk IP/MPLS Backhaul Eth/Nwk Core/MEC
- 4. Page 4 The current IP solution yields huge bandwidth waste for mMTC, UCBC, HCS, and Short Texts 𝐼𝑃 𝐵𝑎𝑐𝑘ℎ𝑎𝑢𝑙 𝑡𝑟𝑎𝑛𝑠𝑝𝑜𝑟𝑡 𝑂𝐻 % = (𝑃𝑎𝑡ℎ𝑜ℎ) (𝑆𝑡𝑑ℎ𝑑𝑟 + 𝑃𝑎𝑡ℎ𝑜ℎ + 𝑈𝑠𝑒𝑟 𝐷𝑎𝑡𝑎) 0 10 20 30 40 50 60 70 80 90 100 OH 1-hop OH 4-hop OH 10-hop OH 20-hop 4-byte: Overhead in % of total length MPLS SR-MPLS SRv6 MPLS bytes MPLS-SR bytes SRV6 bytes IPv6 Encap 40 SRH header 8 Transport Labels 4 to 12 ServiceLabel 4 ServiceLabel 4 Outer IPv4 (for GTP) 20 Outer IPv4 (for GTP) 20 ServiceSID 16 UDP Hdr 4 UDP Hdr 4 UDP Hdr 8 GTP 12 GTP 12 GTP 12 Inner User IP 20 Inner User IP 20 Inner User IP 20 User Transport 4 User Transport 4 User Transport 4 User Payload 4 to 1200 User Payload 4 to 1200 User Payload 4 to 1200 16 x SID count (upto 30) transport SID Transport Labels 4 x SID count (upto 30) 44-52 44-160 100-564 Extra Header ❖ MPLS-SR and SRv6 overheads go up with the number of hops ❖ Protocol efficiency with regards to small packets is very low, and can be as low as below 10% • Connected Industries and Industrial Control: Control Command (1 byte) + Optional User Data (0-4 bytes) • Cloud Driving: Instruction Category (4 bits) + Instruction (12 bits) + Optional User Data (0-4 bytes) • Short Message: Hello (5 bytes), how are doing (13 bytes), … (usually small, often up to 40 bytes in a short text) mMTC Requirements: 5G: 1 million devices per square km => B5G/6G: 10 million devices per square km Backhaul Bandwidth Efficiency: Huge IP/MPLS backhaul transport overhead so that over 80%, and even 90% bandwidth could be wasted
- 5. Page 5 The current IPv4/v6 solution simply can’t fully support uRLLC and RTBC Latency (ms) Packet Loss Ratio 0 100 -3 10 -4 10 -5 10 -6 10 -7 10 -8 10 -9 10 -10 10 References: ITU-T Focus Group on Network 2030 Deliverables 3GPP TS 22.261 V15.5.0 (2018-07) 3GPP TS 22.261 V17.3.0 (2020-07) 3GPP TR 22.804 V16.2.0 (2018-12) 3GPP TS 22.104 V17.3.0 (2020-07) better better 50 10 1 20 Electricity Distribution Holographic Teleport Remote Control Tactile Internet Intelligent Transport Robots High Voltage Electricity Distribution Industrial Automation • Packet can be lost • If retransmitted, the latency is tripled • No guarantee for precise KPI App (user) TCP (user) IP (user) GTP-U (S1) UDP (Network) IP (Network) IP/MPLS/SRv6 Backhaul Ethernet/PPP The uRLLC KPI is mission critical and sometimes life critical for 5G/B5G services, but IPv4/IPv6 cannot guarantee KPI. We need a way to evolve IPv4/v6 for uRLLC and RTBC
- 6. 6 Existing Addressing • Most Popular L3 Addresses – IPv4 and IPv6 • Many Others – Location + ID Separation • LISP (IETF RFC 6830) – Digital Object Architecture (DOA) • DOI (https://www.cnri.reston.va.us/activities.html) – Service ID • draft-jiang-service-oriented-ip-03 – ICN, NDN, CCN • Names and/or IDs for Information and/or Content – ID: E.g., DeviceNet ID (IEC 62026-3) – NSAP (ISO/IEC 8348) – Mixed Addresses • Mixture of MAC, IP, and Naming in Profinet (IEC 61158 / IEC 61784- 1 and IEC 61784-2) – Flexible Addressing System • Variable-length addresses (ACM Sigcomm NEAT 2019) 1) Designed for general connectivity 2) “One size fits all”, but in reality one size may not fit all nicely 3) Prone to address spoofing because of its well- known global format 4) Privacy infringement and surveillance made easier through IPv6 homogenization everywhere 5) Internet Consolidation (or Concentration) ISOC 2019 Global Internet Report 6) Internet Ossification Keynote Speech in ITU-T FG Network 2030 London Meeting 7) Some proposals repurpose components of IPv6 addresses or overload their semantics 8) Address optimization and compression are being proposed, especially for IoT, Industrial IoT e.g. IETF WGs ROLL, 6TISCH, 6LO, etc Analysis of IPv4 and IPv6
- 7. 7 As a matter of fact, for private or limited networks, we don’t have to use globally-formatted network addresses! This is so true for industrial networks (manufacturing, gas pipeline, mining) Source Destination Data 192.168.100.101 192.168.500.200 Hello 192.168.200.* 192.168.500.* 192.168.300.* 192.168.100.* R1 R3 R2 R5 Industrial Network Common Prefix: 192.168.0.0/16 192.168.100.101 192.168.500.200 A B Full Address in Packet from A to B 4768 330-4768 (408) 330-4768 +1 (408) 330-4768 4419 330-4419 (408) 330-4419 +1 (408) 330-4419 ▪ No need to dial the full number 1 408 330 4419 ▪ Dial what is needed depending on where the receiver is P1 P2 Address Type Source Destination Data Short Address 100.101 500.200 Hello Short Address in Packet from A to B OSPF, ISIS No Change BGP No Change No Change RIB FIB The common prefix is truncated PLC ❖ This idea can extend to other contexts and topologies to form flexible variable-length addressing systems ❖ It can be combined with mix and match • Wasteful • Easier to attack • Economical • More Secure
- 8. 8 Source Address Destination Address Source Type Source Address Destination Type Destination Address User Data User Data ❖ Free-Choice Addressing ❖ Mix and Match ▪ It provides freedom to network operators and application developers to choose the most effective addressing system for their domains and applications. ▪ It does not dictate the use of addressing systems. ▪ When a new addressing system is introduced, there is no need to define a newer IP again - A lesson learnt from IPv6 ▪ Existing addressing systems are optimized where possible, and new addressing systems are permitted where necessary. ▪ Industrial domains may customize their addressing systems. For example, If an OT network has only a few thousands of terminals, it can use short versions of IP addresses. ▪ Compare with postal mails: The sender has a Japanese address and the receiver has an English address ▪ Mix-and-Match is powerful when converging IT networks and OT networks. ▪ Example: An IPv6 device can talk with a LISP terminal ▪ Example: An IPv4 PLC controller can talk with a Profinet Class B terminal in geographically different LANs Address Type • Protection of Industrial Networks • Energy-Efficiency for Small Devices of mMTC and HCS • The freedom and autonomy of networks and domains Why do we need new and more addressing schemes for private/limited networks? • Accommodation of Connection for Unconnected Industrial Machines • Integration of Terrestrial Networks and other Networks What can we use as other options for addressing schemes?
- 9. Page 9 Now that IPv4/v6 does have some limits, what can we do about them and how? Answer: An incremental evolution to expand the scope of its applicability for future applications and services Header User Payload KPI (latency, packet loss, etc) Sender’s Intent In-Band OAM and Telemetry Network Programmability IPv4/IPv6 Header Contract User Payload Addressing Evolution Payload Evolution Evolved Header Contract Evolved User Payload Flexible Addressing System Geography-Based Addressing Integration of Satellite and Terrestrial Networks Holographic Teleport Holographic Type Communications Qualitative Communications Entropy-Based Communication Evolution into Modern Courier-like Datagrams References: (1) New IP: A Data Packet Framework to Evolve the Internet, IEEE HPSR 2020, 2020 (2) New IP: Enabling the Next Wave of Networking Innovation, in Design Innovation and Network Architecture for the Future Internet: Computer Science & IT Books | IGI Global (igi-global.com)
- 10. 10 Thank You