CS434/534: Topics in Networked (Networking) Systems

Transcription

1 CS434/534: Topics in Networked (Networking) Systems Improve Wireless Capacity; Programmable Wireless Networks Yang (Richard) Yang Computer Science Department Yale University 208A Watson

2 Outline Recap Wireless background Frequency domain Modulation and demodulation Wireless channels Wireless PHY design Wireless MAC design (one hop) wireless access problem and taxonomy wireless resource partitioning dimensions media access protocols From single-hop MAC to multi-hop mesh wireless mesh network capacity maximize mesh capacity

3 Admin. PS2 due: Tuesday at 11:55 pm Please stop by Geng s office at any time between now and Tuesday deadline Project first check point: A meeting w/ instructor by Monday, Apr. 3 3

4 Recap: Capacity of Mesh Networks The question: how much traffic can a mesh wireless network carry, assuming an oracle to avoid the potential overhead of distributed synchronization (MAC)? Why study this question: learn the fundamental limits of mesh wireless networks, separate the spatial reuse perspective and system design perspective, gain insight for designing effective wireless protocols. 4

5 Wireless Capacity & Capacity Improvements Multiple transceivers Radio interface constraint a single half-duplex transceiver at each node lt å Reduce L h( b) WT n b= 1 2 rate*distance capacity: Approx. optimal l L Reduce interf. area Interference constraint transmission successful if there are no other transmitters within a distance (1+D)r of the receiver 2 16WT lt h( b) h ( rb ) 2 b= 1 h= 1 pd åå 8 W p D n Increase W 5

6 Outline Recap Wireless background Frequency domain Modulation and demodulation Wireless channels Wireless PHY design Wireless MAC design (one hop) From single-hop MAC to multi-hop mesh understanding: wireless mesh network capacity design: maximize mesh capacity reduce L

7 Change Traffic Pattern: Reduce L Reduce L => make communications local node placement: change the demand patterns (thus L) e.g. base stations/access points with high-speed backhaul BS1 infrastructure BS2 F A B C D E T S 7

8 Change Traffic Pattern: Reduce L Reduce L => make communications local by being patient and wait in a mobile network 8

9 Outline Recap Wireless background Frequency domain Modulation and demodulation Wireless channels Wireless PHY design Wireless MAC design (one hop) From single-hop MAC to multi-hop mesh understanding: wireless mesh network capacity design: maximize mesh capacity reduce L MIMO: Use multiple antennas

10 Reduce Interference Footprint Antenna design: steered/switched directional antennas A B C A B C D D Exercise: redo capacity derivation if limited interference. 10

11 Multiple Input Multiple Output (MIMO) 4x4 MIMO LTE Kindle Fire HD 11

12 MIMO Basics 1 x 1 h 11 y 1 1 h 12 2 h 21 y 2 2 h 22 x 2 y y 1 2 = = h h x x Solve two variables from two equations. h h x x

13 Using MIMO for more Concurrency: Motivation No Transmission in current n Assume tx1 is sending to rx1 Can tx2 transmit in using carrier sensing? 13

14 MIMO Benefit: Concurrency using Interference Nulling h h21 h tx2: for every symbol q, transmits q on first antenna and aq on second antenna. interference at rx1: ( a h + h ) q if tx2 picks = - a NO interference at rx1. h h

15 Problem - rx2 hears p from tx1 h h21 h Can rx2 decode? 15

16 Decoding at rx2: Observation h h21 h for different symbols p from tx1, the received signal at rx2 moves along vector! y h! tx1 æh = ç èh 13 ö p ø = h 12! tx1 p Perp. of tx1 space - rx2 can estimate channels h12, h13 from preamble 16

17 Decoding at rx2: y Removing tx1 signal by Projection! = æ ç è h h 13 ö p ø = h 12! tx1 p h h21 h rx2 projects received signal orthogonal to h! tx1 projection space 17

18 Decoding at rx2: Projection Details h h21 h rx2 picks w2 and w3: w 2 * h 12 + w 3 * h 13 = 0 to compute w * + y 2 y2 w3 * 3 projection space 18

19 Decoding at rx2: Projection Details h h21 h w 2 * h 12 + w 3 * h 13 = 0 => w 2 y 2 = [ w ( h w 3 w 22 3 ( h y + ah ) + ah 33 )] q Summary: MIMO allows concurrent transmissions. 19

20 Problem of Only Nulling If only nulling, tx3 cannot transmit Assume both tx1 and tx2 are transmitting. 20

21 Solution: MIMO using Interference Alignment Key idea: rx2 ignores interference from tx1 by projection. If tx3 aligns tx3 -> rx2 interference along the same direction as that of tx1 -> rx2, then rx2 can remove it too. Assume both tx1 and tx2 are transmitting. 21

22 MIMO with Nulling and Alignment tx3 picks a, b, g rx2 sees: Because rx2 projects to orthogonal to interference from tx3 to rx2 h! tx1, no 22

23 Outline Recap Wireless background Frequency domain Modulation and demodulation Wireless channels Wireless PHY design Wireless MAC design (one hop) From single-hop MAC to multi-hop mesh understanding: wireless mesh network capacity design: maximize mesh capacity reduce L MIMO: Use multiple antennas Cognitive radio: use unlicensed spectrum

24 Spectrum Allocation Chart 24

25 Unlicensed Spectrum Opportunity: unlicensed spectrum is large and has low utilization US unlicensed freq: G M G G (200 mw) (1 w) (4w) 25

26 Problem of Using Unlicenced Unlicensed spectrum may have occupants and is fragmented Zigbee a Others Unlicensed Spectrum Requirement: Coexistence with dynamic and unknown narrowband devices in the unlicensed spectrum 26

27 Cognitive Aggregation Cognition: Detect unoccupied bands Aggregation: Weave all unoccupied bands into one link Wideband Unlicensed Spectrum Zigbee a Others

28 Research Issues How to detect available frequency bands? How to operate across chunks of noncontiguous frequencies? How do sender and receiver establish communication when their perceived available frequency bands differ?

29 Outline Recap Wireless background Frequency domain Modulation and demodulation Wireless channels Wireless PHY design Wireless MAC design (one hop) From single-hop MAC to multi-hop mesh Wireless networking software framework

30 Big Picture Applications Wireless/Mobile Application Development Framework Foundational Services: Communications, Location, Service Discovery, UI/Media, Power Management, Security 30

31 Overview Wireless/mobile software development framework for mobile wireless applications is a quite large topic We will cover both examples and basic principles SDR GNURadio, SORA, OpenRadio Software OS for sensors TinyOS Mobile J2ME Android, IOS Approach for designing/evaluating each software development framework: Focus on the key concepts introduced by each framework 31

32 Outline Recap Wireless background Frequency domain Modulation and demodulation Wireless channels Wireless PHY design Wireless MAC design (one hop) From single-hop MAC to multi-hop mesh Wireless networking software framework Software defined radio (SDR)

33 Software Defined Radio Instead of hardware-centric radio, making wireless networking systems softwarecentric. Key challenges Handle large data volume Provide hard deadline and accurate timing control Provide easy to use, compose signal processing blocks 33

34 Outline Recap Wireless background Frequency domain Modulation and demodulation Wireless channels Wireless PHY design Wireless MAC design (one hop) From single-hop MAC to multi-hop mesh Wireless and mobile software framework Software defined radio (SDR) GNURadio

35 GNURadio Hardware Arch Hardware Frontend Host Computer RF Frontend (Daugtherboard) ADC/DAC and Digital Frontend (USRP) GNU Radio Software

36 Basic Software Concepts q Block q Flow graph

37 Basic Software Concepts q gr_basic_block (name, in/out signature, msg queue) q see 3.4/classgr basic block.html q gr_block (leaf block; key functions forecast/general_work) q see 3.4/classgr block.html q gr_hier_block2 (container block; key functions: connect/disconnect/lock/unlock) q gr_top_block (flow graph; start/stop/wait) What is the design pattern?

38 Software/Execution Model q Software model q q q C++ o o Signal processing blocks Certain routines also coded in assembly Python o o o Application management (e.g., GUI) Flow graph construction Non-streaming code (e.g., MAC-layer) Execution model q Python thread for each top_block Python Application development Flow graph construction C++ Signal processing blocks Discussion: benefits/issues of the hybrid software structure?

39 Summary: GNURadio Interesting/key software design techniques you learned from GNURadio? Composite pattern to build a hierarchy of blocks Define gr_block as a reusable base so that defining a new block is typically simple: overwrite general_work and forecast Internal scheduler to orchestrate data flows among blocks Hybrid software system (Python/C++) 39

40 Outline Recap Wireless background Frequency domain Modulation and demodulation Wireless channels Wireless PHY design Wireless MAC design (one hop) From single-hop MAC to multi-hop mesh Wireless and mobile software framework GNURadio SORA

41 Sora: Goal 41

42 Sora: Hardware (RF) 42

43 Sora: Hardware (RCB)

44 Sora: Software Efficient impl of blocks using LUT/SIMD Utilizing multi-core for streaming processing Dedicated core for real-time support

45 Sora Software: Acceleration (LUT) Utilize cache (LUT) Extensive use of lookup tables (LUT): trade memory for calculations; still well fit into L2 cache Applicable to more than half of the common alg; speedup ranges 1.5x-22x Example:

46 Sora Software: Acceleration (SIMD) Utilize data para. (SIMD) Modern CPUs support wide-vector SIMD ext. Intel SSE 128 packed vector, allowing 8 x 16bit ops Applicable to many PHY alg. w/ speedups (1.6x- 50x) Example: FIR

47 Sora Software: Acceleration

48 Sora Software: Multicore Streaming Partition and schedule PHY processing across cores Interconnecting sub-pipeline w/ light-weight sync. FIFOs Static scheduling of processing modules in PHY pipeline

49 Summary: Sora Interesting/key software design techniques you learned from Sora? 49

50 Outline Recap Wireless background Frequency domain Modulation and demodulation Wireless channels Wireless PHY design Wireless MAC design (one hop) From single-hop MAC to multi-hop mesh Wireless and mobile software framework GNURadio SORA TinyOS

51 Design Goal A free and open source component based operating system and platform targeting wireless sensor networks (WSNs) Example app Environment monitoring, e.g., measure temperature, lighting values/events periodically transmit measurements/events to a base station forward data for other nodes that are out of range of the base station 51

52 Hardware 1.5 x 1.5 Assembled from off-the-shelf components 4Mhz, 8bit MCU (ATMEL) 512 bytes RAM, 8KB ROM Devices serial port temperature sensor & light sensor 900Mhz Radio (RF monolithics) ft. range LED outputs 52

53 Schematic Diagram of a Mote 53

54 Requirements on Software Dev. Framework Flexible configuration of attached devices Small foot print devices have limited memory and power resources 54