640 likes | 795 Views
Beyond (Basic) Cellular Networks: Multi-Hop/Meshed, Ad-Hoc, DTNs, White Space, Wireless Cloud …. Shivkumar Kalyanaraman shivkumar-k AT in DOT ibm DOT com http://www.shivkumar.org Google: “shivkumar ibm rpi”.
E N D
Beyond (Basic) Cellular Networks: Multi-Hop/Meshed, Ad-Hoc, DTNs, White Space, Wireless Cloud … Shivkumar Kalyanaraman shivkumar-k AT in DOT ibm DOT com http://www.shivkumar.org Google: “shivkumar ibm rpi” Based in part upon slides of Bhaskaran Raman, Kameswari Chebrolu, Mihail L. Sichitiu, Hari Balakrishnan …
Outline • Multi-hop (meshed) networks • Dynamic Ad-Hoc Networking & Weak State Routing • Delay/Disruption Tolerant Networks (DTNs) • Opportunistic access / offload • Software Radio & Wireless Network Cloud • Cooperative MIMO & other cooperative techniques • White Space Networking
Rate 4G 802.11b WLAN 3G Other Tradeoffs: Rate vs. Coverage Rate vs. Delay Rate vs. Cost Rate vs. Energy 2G 2G Cellular Mobility • Scarce bandwidth • (10-100 MHz/operator) 2. Spectral Efficiency: MHz -> Mbps (signal to noise ratio is key!) 3. Tradeoffs: Rate vs X (no free lunch!) TodayWith femto cells & MIMO antennas Recall: Wireless: A Short Technical summary Wireless networks are designed to maximize spectral efficiency, support mobility, coverage, and Quality-of-Service under severe spectrum/bandwidth constraints Wireless IT convergence
2km CDMA BS frequency time OFDM 50m BS Spectrum Scarcity & Solutions • CDMA/OFDMA & Scheduling: • Spread spectrum in time-domain (CDMA) or frequency domain (OFDMA) • Statistical multiplexing of time-frequency • Dynamic tradeoff of power/interference: i/f limited • More signal processing (3G/3.5G): voice/data • Pico/Femto Cells & (WiFi) Overlays/ Offload: • Smaller Cells, lower power • Indoor access characteristics: offload onto Wifi or Femto cells • SON management • Macro cell overlays • Frequency Reuse: • Static spectrum sharing; TDM within each cell (GSM) • Adjacent cells use different frequency • Widely used for voice communication (GSM) • Coverage radius reduced and low spectral efficiency • MIMO & CoMP: • Multiple antennas: Spatial degrees of freedom • Collaborative MIMO (CoMP) to manage inter-cell interference • Intensive signal processing, channel estimation, BS coordination • Easier with BS Pools/Cloud
Underlay Occupied Occupied Occupied spectrum spectrum spectrum UWB UWB Underlay Spectrum resource Spectrum resource Spectrum Scarcity & Solutions (contd) Spectrum Scarcity • Multi-hop/Meshed Networks / 60 GHz/FSO: • Smaller hops: O(sqrt(N)) capacity increase • Multi-route diversity; resilient transport protocols; routing advances • Backhaul or shared spectrum • Spectrum at higher reaches (60GHz, FSO) less regulated; dense spatial reuse • Space-Time-Frequency Shifting of Workloads: • Mobile mini-base station fleet • Opportunistic access to smaller cells (associations for ~10s) • Multi-homed mobile devices • Time-shifting of content delivery; sophisticated traffic shaping at peak times. • Delay / disruption-tolerant/ad-hoc networks • White space/Cognitive Radio/Opportunistic Spectrum Access: • Using white space of spectrum opportunistically • Dynamic spectrum scheduling and management • Need complex technologies for detecting the white spectrum space and management policies • Fit into cellular model TBD • UWB (Underlay): • Using ultra wide band spectrum without disturbing the occupied users • Need to control the transmitted power below noise level to avoid interference • Generally for short range transmission
Taxonomy Wireless Networking Single Hop Multi-hop Infrastructure-based (hub&spoke) Infrastructure-less (ad-hoc) Infrastructure-based (Hybrid) Infrastructure-less (MANET) 802.11 802.16 802.11 Bluetooth Cellular Networks Car-to-car Networks (VANETs) Wireless Sensor Networks Wireless Mesh Networks
Mesh vs. Ad-Hoc Networks • Multihop • Nodes are wireless, possibly mobile • May rely on infrastructure • Most traffic is user-to-user Wireless Mesh Networks Ad-Hoc Networks • Multihop • Nodes are wireless, some mobile, some fixed • It relies on infrastructure • Most traffic is user-to-gateway
Mesh vs. Sensor Networks Wireless Sensor Networks Wireless Mesh Networks • Bandwidth is generous (>1Mbps) • Some nodes mobile, some fixed • Normally not energy limited • Resources are not an issue • Most traffic is user-to-gateway • Bandwidth limited (tens of kbps) • In most applications, fixed nodes • Energy efficiency is an issue • Resource constrained • Most traffic is user-to-gateway
Goals: variety of apps, QoS, scalable operation (100-200 nodes)
Meshed Networks: Issues • The “link” abstraction may or may not hold depending upon the type of links designed (directional vs omni) • Using 802.11 MAC for multi-hop does not work • Usually meshes are ~3 hops diameter • TDMA / OFDMA extensions to handle 802.11 issues • Per-hop losses: overcome via loss-tolerant TCP or multi-path LT-TCP • Community meshes havent been too successful • Niches in smart grids etc.
Ch-1 3 2 2 1 1 Ch-1 Ch-1 2 Ch-2 1 3 4 3 4 Ch-1 Ch-2 Chain bandwidth = B User bandwidth = B/2 User bandwidth = B MAC – Multichannel • Increases network capacity B = bandwidth of a channel
GW GW GW MAC – MultichannelStandard MAC – Multiple Radios • A node now can receive while transmitting • Practical problems with antennas separation (carrier sense from nearby channel) • Optimal assignment – NP complete problem • Solutions • Centralized • Distributed
GW GW GW MAC – MultichannelCustom MAC – Multiple Radios • Nodes can use a control channel to coordinate and the rest to exchange data. • In some conditions can be very efficient. • However the control channel can be: • an unacceptable overhead; • a bottleneck;
Routing for Meshes and MANETs • Routing consists of two fundamental steps • Data plane: Forwarding packets to the next hop (from an input to an output interface in a traditional wired network) • Control plane: Determining how to forward packets (building a routing table or specifying a route) • Forwarding packets is easy, but knowing where to forward packets (especially efficiently) is hard • Reach the destination • Minimize the number of hops (path length) • Minimize delay • Minimize packet loss • Minimize cost
MANET vs. Traditional Routing • Every node is potentially a router in a MANET, while most nodes in traditional wired networks do not route packets • Topologies are dynamic in MANETs due to mobile nodes, but are relatively static in traditional networks • Channel properties, including capacity and error rates, mostly static in traditional networks, but vary in MANETs • Routing in MANETs could consider both Layer 3 and Layer 2 information: L2 can indicate connectivity and interference • Interference is an issue in MANETs, but not in traditional networks • Channels can be asymmetric with some Layer 2 technologies • Traditional routing protocols for wired networks do not work well in most MANETs: too dynamic
Types of MANET Routing MANET Routing Protocols Proactive Reactive Hybrid Example: OLSR Example: AODV
Common Features • MANET routing protocols must… • Discover a path from source to destination • Maintain that path (e.g., if an intermediate node moves and breaks the path) • Define mechanisms to exchange routing information • Reactive protocols • Discover a path when a packet needs to be transmitted and no known path exists • Attempt to alter the path when a routing failure occurs • Proactive protocols • Find paths, in advance, for all source-pair destinations • Periodically exchange routing information to maintain paths
Geographic Routing Geographic Routing: Compared to topology-based routing schemes, geographic routing schemes forward packets by only using the position information of nodes in the vicinity and the destination node. Thus, topology change has less impact on the geographic routing than other routing protocols. Early geographic routing algorithms are a type of single-path greedy routing schemes in which packet forwarding decision is made based on the location information of current forwarding node, its neighbors, and the destination node. However, all greedy routing algorithms have a common problem, i.e., delivery is not guaranteed even if a path exists between source and destination.
Challenges in Routing for Large-scale & Dynamic Networks • Routing table entries: “state” = indirections from persistent names (ID) to locators • Due to dynamism, such indirections break • Problematic in two dimensions • Dynamism/mobility => frequent update of state • Dynamism + large scale => very high overhead, hard to maintain structure • Proposed solution: • Probabilistic and more stable state WEAK STATE • Use of unstructured methods Node Mobility Number of Nodes
Strong State Deterministic Requires control traffic to refresh Rapidly invalidated in dynamic environments Weak State Probabilistic hints Updated locally Exhibits persistence Weak State: A New Type of State
INSTALL REMOVE REFRESH Hard, Soft and Weak State s r A with probability B A B A Time elapsed since state installed/refreshed Confidence in state information () Hard State Soft State Weak State Weak State is natural generalization of Soft State
An Instance of Weak State • The uncertainty in the mappings is captured by locally weakening/decaying the state • Other realizations are possible • Prophet, EDBF etc… SetofIDs GeoRegion {a,b,c,d,e,f} Probabilistic in terms of scope Probabilistic in terms of membership
Weak State Routing for MANETs Random Directional Walk (RDW) • RDW used to announce location information (“put”) and forward packets (“get”)
Dissemination/Proactive Phase: (put) • When a node receives a location announcement, it • creates a ID-to-location mapping • aggregates this mapping with previously created mappings if possible C B A
Forwarding Packets (get) A S B WSR involves unstructured, flat, but scalable routing ; no flooding ! C E D Forwarding decision: similar to longest-prefix-match. “strongest semantics match” to decide how to bias the random walk. Details in ACM Mobicom 2007 paper
Packet Delivery Ratio – Fixed Density WSR always achieve high delivery ratio GLS works fine at low mobility but fails to maintain structure at high dynamism OLSR delivers only a small fraction even at low dynamism
Control Packet Overhead OLSR overhead increases exponentially GLS works fine at low mobility but requires superlinearly increasing overhead to maintain structure at high mobility
Focus on eliminating the uncertainty between congestion loss and all other reasons Many approaches developed for single-hop wireless systems Snoop I-TCP M-TCP End to end SACK Explicit error notification Explicit congestion notification (e.g. RED) New solutions for multi-hop Loss-Tolerant TCP Multi-path LT-TCP (MPLOT) Transport: TCP Solutions
Maximum Goodput Missing Goodput! Loss-Tolerant TCP (LT-TCP) vs TCP-SACK
High Delay/Jitter Low Capacity Lossy Single path: limitedcapacity, delay, loss… Time Network paths usually have: • low e2e capacity, • high latencies and • high/variable loss rates.
Low Perceived Loss High Perceived Capacity Scalable Performance Boost with ↑ Paths Low Perceived Delay/Jitter Idea: Aggregate Capacity, Use Route Diversity!
Multi-path LT-TCP (ML-TCP): Structure Socket Buffer Map pkts→paths intelligently based upon Rank(pi, RTTi, wi) Per-path congestion control (like TCP) Reliability @ aggregate, across paths (FEC block = weighted sum of windows, PFEC based upon weighted average loss rate) Note: these ideas can be applied to other link-level multi-homing, Network-level virtual paths, non-TCP transport protocols (including video-streaming)
DTN Examples Delay and Disruptions are first-class issues End-to-end path may never exist at any instant in time, but may emerge only over time
Overview of Routing Issues for DTNs • State vs stateless Routing: • Stateless: completely depend upon mobility, local storage at nodes & replication/coding • Eg: “Spray-and-wait” , “Spray-and-focus” • Scaling challenges. • Stateful: How to maintain useful state info despite disconnections. • Weak state can again help (eg: WSR-D protocol), with “osmosis” of state across connectivity clusters. • Simple situations such as “data mule” (getting e-mail from a village, or synchronizing photos) involve 1-hop DTN routing etc. • Vehicular DTNs (eg: for an entire city) to provide useful complementary communication services to cellular: not yet fully solved. • Interesting small-scale testbeds: DieselNet (UMass)
Opportunistic Offload via Small Cells Opportunistic traffic offload Eg: Aruna Balasubramanian, Ratul Mahajan, Arun Venkataramani Augmenting Mobile 3G Using WiFi: Measurement, Design, and Implementation In Proceedings of ACM MobiSys, San Francisco, USA, June 2010. Vladimir Bychkovsky, Bret Hull, Allen K. Miu, Hari Balakrishnan, Samuel Madden, “A Measurement Study of Vehicular Internet Access Using In Situ Wi-Fi Networks,” Proceedings of ACM Mobicom 2006, Los Angeles, 2006. (best paper)
Wi-Fi Is Everywhere (in developed urban markets) Images from WiGLE.net and CarTel
Opportinistic Offload: The Opportunity • Today: • Broadband connections are often idle • 65% of on-line households have Wi-Fi • What if … • … home users open up their APs … • … and share/sell the spare bandwidth? • Cellular complement for mobile users: • Messaging (multimedia, e-mail, text) • Location-aware services • Mobile sensor networks (e.g. MIT project CarTel )
Wi-Fi For Mobile Messaging • Wi-Fi cells are smaller than cellular cells • Is density sufficient? Are connections too short? • Organically grown, unplanned deployments • Uneven densities, AP churn, unpredictable • Back-of-the-envelope: • 55 km/hour: ~15 meters/s • ~150 meter AP coverage [Akella’05] • ~10 sec connectivity • What about connection overhead? • scan, associate, get IP, etc. • Current stacks too slow
CarTel Expt: Bytes Uploaded Per Connection Non-trivial amount of data: Median: 200 KBytes per connection Mean: 600 KBytes Fraction of connections Consistency check: 600 KBytes / 24 sec = 25 KBps Bytes received on server (KBytes)
The Future of Software Radio: Wireless Network Cloud Parul Gupta, Smruti Sarangi, Shivkumar Kalyanaraman [IBM Research – India] Zhen Bo Zhu, Lin Chen, Yong Hua Lin, Ling Shao [IBM Research – China]
Edge gateway ManagementServer SMS/MMS SMS/MMS BS WAP GW IMS Content Service BS Web Service Edgegateway Billing BS BS PSTN 2G-3G wireless network architecture 4G Wireless Network over Wireless Network Cloud Cloud of Wireless Access Network + Core Network Access Network Core Network Service Network Mobile switch center BS cluster Radio network controller Service support node Radio network controller Gateway Internet BS cluster