Need for Improved Loss Tolerance. K. K. Ramakrishnan. Thanks to: Robert Miller(AT&T), Vijay Subramanian and Shiv Kalyanaraman (RPI)

Transcription

1 4G Wireless Networks Need for Improved Loss Tolerance K. K. Ramakrishnan AT&T Labs Research, NJ Thanks to: Robert Miller(AT&T), Vijay Subramanian and Shiv Kalyanaraman (RPI)

2 Adaptive Wireless: Goal is to Make Wireless As Good As Wired User Rate Adaptive modulation and coding exploits previously inaccessable capacity enabling higher rates Shannon Adaptive Modulation and Error Correction Coding Provide Higher Throughputs if Propagation Allows Multiple Antennas and Space-Time Coding Combined with Better Cell Propagation Predictability Provide Wire-Like Performance 4G Technologies e.g. MIMO, Small-Cells (802.11n) Limit Potential Improvement (3G, ) 2.5G Systems Users in this region could get some connectivity rather than no service Signal to Noise Ratio Fixed modulation and coding provides the same rate to all users, even if their links could provide higher rates Page 2

3 Reach-Data rate Tradeoff 100 Higher Rate, Lower-Speed Mobility Peak Data Rate Megabit ts per Seco ond/user 10 1 UWB Bluetooth Zigbee PANs.1 2.4GHz and UWB Zigbee (US) Zigbee (Europe) 4G H/S Wireless LAN 2.4 & 5 GHz Unlicensed 4G Wireless NAN 2.4 & 5 GHz 3G/MAN Fixed or Pedestrian 3G/MAN Mobile 2.5G Mobile/Pedestrian 3G/ Wireless Various Bands Wider Area, Higher-Speed Mobility 2/2.5G Wireless 800 MHz, 2 GHz 10 feet 100 feet 1 mile 10 miles Reach Page 3

4 Cell Coverage Area Trends Cell Radius (Feet) 1,000,000 10,000 mi 2 Maritime Mobile HF Radio Service (~300 mi) Increased Bandwidth Demand/User Battery/Dissipation Device Constraints Moore s Law Radios Increased Edge Intelligence Distributed Control Techniques 1G Macrocellular Systems (~8 mi) 100,000 2G Cellular Expanded Service MJ-MK (~4 mi) Mobile 10,000 Telephone Metroliner (~60 mi) Train The 2G Telephone Sweet Spot (~15 mi) 1, Watts 10 Watts 700 mi 2 Mobile/ 2.5G Microcells (~2 mi) PCS Microcells (~0.5-2 mi) The 3G/WiMax Sweet Spot WNAN/LAN Nanocells (~ mi) Year 1 Watt 100 mw.01 mi 2 30 mw The 4G Sweet Spot Portable Maximum Power Output Page 4

5 Architecting a 2G-3G-4G Wireless Network Infrastructure AT&T VoIP Infrastructure Gateway BE AT&T Common BB Backbone BE BE PSTN IS-41 TDM Voice BB Access Network Home 4G 4G Hot Zones Enterprise 4G MSC Cellular Network MSC Municipal Network BS BS BS BS BS BS Home 4G 4G Hot Zones Enterprise 4G Page 5

6 Page 6 How Might A Wireless Deployment Look?

7 The Muni Network Concept Gets Legs The number of municipalities operating, installing, or planning Wi-Fi networks is growing rapidly, driven by User acceptance/use of Wi-Fi Large number of devices already equipped (instant customer base) Availability of low-cost networking systems (including mesh) Ability to transport Ethernet-like throughputs Desire to project Cybercity image Digital Divide amelioration Improvement in public service communications capabilities Leverage existing municipal infrastructure and fiber Revenue opportunities/new business models Ability to raise bond capital for infrastructure Page 7

8 Fiber, 4G, and Multi-Tier Wireless: A NanoNet 3G / Metro Area Wireless Hub (PTMP Fixed Service to Businesses, MDUs, EcoPoles w/o fiber availability) Fiber Ring Fiber PON Local PON Hub City Fiber Hub Eco- Pole AP NAN Cell LAN Indoor Window Cell Bridge Eco- Pole AP EcoPole AP Page 8

9 NanoNet Indoor Coverage Using Window Bridging 3G / Metro Area Wireless Hub (PTMP Fixed Service to Businesses, MDUs, EcoPoles w/o fiber availability) Fiber Ring Fiber PON Local PON Hub Eco- Pole AP City Fiber Hub NAN Cell LAN Indoor Window Cell Bridge Eco- Pole AP Page 9

10 Page 10 Solving Outdoor-to-Indoor Penetration: The Window Bridge

11 Motivation for our work on LT-TCP Dense wireless deployments in urban areas/ high rises will cause disruptions/ burst errors due to interference Protocols need to be loss tolerant and provide reliability l Especially as we move to multi-hop wireless environments Divide the burden of reliability between link and transport layers Keep Residual Loss Rate low; Delay small; Link and Transport Layer Goodput high Page 11

12 TCP-SACK Performance under Lossy Conditions Sharp drop-off in performance with PER (degrades beyond an error rate of 5% PER) Performance is poorer as combination of PER and RTT grows Page 12

13 Goals for our Enhancements to TCP Dynamic Range: Can we extend the dynamic range of TCP into high loss regimes? Can TCP perform close to the theoretical capacity achievable under high loss rates? Congestion Response: How should TCP respond to notifications due to congestion.. but not respond to packet erasures that do not signal congestion? Mix of Reliability Mechanisms: What mechanisms should be used to extend the operating point of TCP into loss rates from 0% - 50 % packet loss rate? How can Forward Error Correction (FEC) help? How should the FEC be split between sending it proactively (insuring the data in anticipation of loss) and reactively (sending FEC in response to a loss)? Timeout Avoidance: Timeouts: Useful as a fall-back mechanism but wasteful otherwise especially under high loss rates. How can we add mechanisms to minimize timeouts? Our Enhancements to TCP: Loss Tolerant-TCP (LT-TCP) Page 13

14 LT-TCP Performance Performance of LT-TCP is much better compared to that of TCP-SACK LT-TCP degrades gracefully (nearly linear degradation with loss rate) Page 14

15 Transport layer performance with loss tolerance across layers Combining Loss Tolerance at the Transport layer with Link layer enhancements allows us to strike a balance in providing the appropriate loss tolerance over a wide range of losses Limiting ARQ at link layer to manage latency Manageable link level residual loss rate Reasonable Goodput even under extreme conditions High Loss Rates: both LL-HARQ+LT-TCP TCP needed to get better performance Low Loss Rates: just LL-HARQ (link) helps to get better performance Page 15