1. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 1 Building Blocks for Mobile Free-Space-Optical Networks Jayasri Akella, Chang Liu, David Partyka, Murat Yuksel, Shiv Kalyanaraman, and Partha Dutta Rensselaer Polytechnic Institute Emails: sri@networks.ecse.rpi.edu, yuksem@ecse.rpi.edu : shiv rpi

2. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 2 Outline q Context and Motivation q Auto-configurable optical antenna design. q Tessellated Spherical Optical Antenna q Auto-alignment Circuit q Mobility Experiment q Simulating Mobile FSO Networks q Results and summary

3. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 3 Bringing Optical Communications and Ad Hoc Networking Together… Mobile communication Auto-configuration Free-Space-Optical Communications (FSO) Ad Hoc Networking Free-Space-Optical Ad Hoc Networks Spatial reuse and angular diversity in nodes Low power and secure Electronic auto-alignment Optical auto-configuration (switching, routing) High bandwidth Low power Directional This paper proposes initial building blocks for this vision…

4. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 4 Current Commercial FSO Point-to-Point Links in dense metros, competing with wires and leased lines Issues: How to achieve link reliability/availability despite weather

5. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 5 Current RF-Based Ad Hoc Networks q 802.1x with omni-directional RF antennas q High-power – typically the most power consuming part of laptops q Low bandwidth – typically the bottleneck link in the chain q Error-prone, high losses

6. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 6 Contributions: Ad-hoc FSO BBlocks q New optical antenna design q Spherical/Honeycomb structure with FSO trans-receiver modules q Potential: logical links function even when antennas are in relative motion. q Auto-configuration circuit that enables physical FSO channel handoff q Integrated with optical antenna design q Simulation models in ns-2 to enable future studies of FSO MANETs q Initial tests suggest need to revisit routing and TCP layer designs

7. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 7 FSO Basics q High-brightness LEDs (HBLEDs) are very low cost and highly reliable components q 35-65 cents a piece, 10 years lifetime q Low power consumption (100 microwatts for 10-100 Mbps!) q 4-5 orders of magnitude improvement in energy/bit compared to RF q Directional => Huge spatial reuse q But…FSO also requires: q availability of unobstructed line-of-sight (LOS) and, q alignment of LOS between the transmitter and the receiver.

8. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 8 LOS Alignment: Optical Antenna Concept q Tessellated spheres with trans- receiver pairs q Line-of-sight (LOS) auto-alignment electronics q Rapid alignment & handoff => enables mobility or sway, while maintaining the logical link. (a)Tessellated Sphere b) Showing a Line of Sight Sphere Tessellated with LED+PD transceivers.

9. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 9 Auto-Alignment Circuit Design q Pilot signal sent q If aligned, signal is fed-back q Feedback signal detection => alignment! q Handoff logical link & transmit data

10. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 10 Alignment Circuit (Contd)

11. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 11 Optical antenna: Multiple Alignment Circuits q Multiple channels => q Connected to a bank of auto-alignment circuits q Eg: 4-circuit block diagram shown below

12. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 12 Optical antenna: Experimental Platform (Contd) q LEDs: high divergence angle q PDs: angular field of view => the LED-PD pair forms a transceiver cone. q The transceiver cone covers a significant volume of 3- dimensional space. q Key: appropriate packing density to cover entire 360 steradian of surrounding space. Tessellated Spherical antennas on stable optical testing platforms

13. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 13 In Action: 4-channel spherical optical antenna Not Aligned: Searching phase to locate an LOS (all channels searching) Aligned: Data Transmission phase (only one channel active)

14. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 14 Mobility Experiment q UDP data transfer between the moving toy train on a circular track and a data-sink at the center of the circle.

15. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 15 Auto-alignment: Search Phase

16. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 16 Auto-alignment: Aligned! Data Transfer Phase

17. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 17 Intensity vs Mobility @ The optical antenna Detector Threshold Not aligned Aligned Denser packing will allow fewer interruptions (and smaller buffering), but more handoffs

18. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 18 Simulation Model: Mobile Ad-Hoc FSO nodes q NS-2 Model: q FSO propagation model (weather effects) q FSO antennas (sphere model) q Additional parameters: q directional normals, transmission and receiving angles q … to assist the propagation model and LOS calculations.

19. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 19 Mobile FSO Simulation (Contd) q Initial proof-of-concept: q 2-D scenario on the XY- plane q Spatial reuse and angular diversity features illustrated. q Single mobile FSO node q Circles around four stationary FSO nodes q Stationary nodes are connected via wired links to a single central node.

20. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 20 q Four experiments, varying: q Speed of mobile node and q Distance of the mobile node from the central node ExperimentMobile Nodes VelocityMobile Nodes Path Radius 11.5 meters/second25 meters 21.5 meters/second35 meters 32.5 meters/second25 meters 42.5 meters/second35 meters Simple Experiments…

21. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 21 TCP sequence numbers in Experiment 1

22. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 22 Experiment 1: Data Transfer in Bursts After LOS discovery

23. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 23 Experiment 2: Higher distance => TCP interactions & lower throughput

24. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 24 Experiment 3: Lower distance, higher speeds TCP affected by higher loss rates & periodic disconnections

25. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 25 Observations q Need dense tessellation and packing. q Need rapid auto-alignment q TCP may be affected with increasing distance and speed. q End-to-end connection is not the same as physical link alignment q Key: q Need to provide either bit-level buffering and/or q Link-layer hybrid ARQ/FEC to mask such losses from TCP q Interactions with transport and network level protocols will need to be studied and optimized… q Ongoing work…

26. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 26 Indoor Ad-Hoc FSO: Music App

27. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 27 Summary q Ad-hoc FSO communication: q Different from pt-pt FSO and ad-hoc RF q Key building blocks: q Optical antenna: tessellated sphere with dense packing of trans-receivers q Auto-Alignment optoelectronic circuit (simple feedback design) q Absence of mechanical parts such as motors or moving mirrors typically used for auto-alignment purpose. q Significant savings in power consumption and improved alignment reliability. q Simple demonstration: optical data transmission between toy train and ground nodes q NS-2 simulation components: q FSO propagation models q Mobile FSO antennas. q Initial simulation: points to need for optimizing interactions w/ transport and network level protocols.

28. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 28 Thanks! : shiv rpi Students: Jayasri Akella, sri@networks.ecse.rpi.edusri@networks.ecse.rpi.edu Dr. Murat Yuksel (post-doc): yuksem@ecse.rpi.eduyuksem@ecse.rpi.edu Chang Liu, c.liu@ee.unimelb.edu.auc.liu@ee.unimelb.edu.au David Partyka, partyd@rpi.edupartyd@rpi.edu Sujatha Sridharan Ps: Online free videos of all my advanced networking classes

29. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 29 Details

30. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 30 Simulation of mobile FSO nodes continued q An FTP session is kept alive between the central node and the mobile node. For our experiments, all wired links are 100 Mbps with 2ms delays and Drop Tail queues, while the FSO nodes are configured to only transmit at 20 Mbps. (20Mbps is just our configuration limitation, and is not a physical limitation as modulation speeds can be in the order of GHz in optical bands) q Initially, the experiment starts with the mobile node and one of the stationary nodes in LOS. q Soon after the session is established, the node moves around the stationary nodes at a constant rate of speed. Routing is performed by ad hoc DSDV routing agents and MAC is facilitated by 802.11 that is already present in NS-2.

31. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 31 Simulation of mobile FSO nodes continued q We can see that using FSO propagation model in the simulation, it is possible to achieve connectivity through mobile FSO communication even with a very small number of transceivers on the spherical optical antenna. q The experiments were configured in such a manner that LOS is not always present, thus showing that connectivity is reestablished when the nodes are back in LOS. This is demonstrated by the periods of inactivity in the utilization graphs and by the plateaus in the TCP sequence number graphs, which is shown in the figure. q The TCP sequence numbers for the other experiments also showed similar behavior, where plateaus exist for connectivity periods. q Furthermore, increase in the TCP sequence numbers imply that: q All simulation components from physical layer to transport layer are setup properly, thereby provides validity of our simulation building blocks. q Transport level good-put can be achieved over a highly variant (i.e. frequent LOS changes) FSO environment.