π Lecture 30 β QoS in WLAN / Mobile IP
π Overview: This lecture addresses the Quality of Service (QoS) limitations inherent in legacy IEEE 802.11 WLANs and introduces the 802.11e standard designed to overcome them through new mechanisms like EDCF and HCF. The second half covers the fundamentals of Mobile IP, enabling seamless mobility for hosts across networks while maintaining their permanent IP addresses.
ποΈ Topics Covered
The lecture begins with QoS limitations of the original 802.11 DCF and PCF modes, then provides an overview of the 802.11e standard including Traffic Categories (TCs) and Access Categories (ACs). It details the Enhanced Distributed Coordination Function (EDCF) and Hybrid Coordination Function (HCF) mechanisms. The second half introduces Mobile IP basics, care-of addresses, the three protocol capabilities (discovery, registration, tunneling), IP-in-IP tunneling, triangle routing inefficiencies, route optimization, and Mobile IPv6 roaming.
π Lecture Summary
Last Lecture
The previous lecture covered types of attacks and the goals of 802.11 security including access control, data integrity, and confidentiality. It discussed the WEP Protocol, WEP Authentication, and security flaws in original 802.11. Additional topics included 802.1x Security with AKM operations using both AS and PSK, and the IBSS Security model.
QoS Limitations of 802.11
DCF (Distributed Coordination Function) only supports best-effort services with no guarantee in bandwidth, packet delay, or jitter. It suffers from throughput degradation under heavy load. PCF (Point Coordination Function) has an inefficient central polling scheme with unpredictable beacon frame delay due to incompatible cooperation between Contention Period (CP) and Contention Free Period (CFP) modes. The transmission time of polled stations is unknown.
Overview of 802.11e
The 802.11e task group formed in September 1999, with the QoS baseline document approved in November 2000. The standard aims to support both IntServ and DiffServ models. New QoS mechanisms include EDCF (Enhanced DCF) and HCF (Hybrid Coordination Function). 802.11e is backwardly compatible with DCF and PCF.
π‘ Why this matters: 802.11e introduced the first standardized QoS framework for Wi-Fi, enabling networks to prioritize voice and video traffic over data traffic.
QoS is realized by introducing Traffic Categories (TCs) . MSDUs are delivered through multiple backoff instances running as virtual stations, each parameterized with TC-specific parameters: AIFS, CWmin, CWmax, and Persistence Factor (PF) . For legacy DCF, AIFS=DIFS, PF=2, and CWmin<15. The new CW calculation: CWnew[TC] = (CWold[TC]+1) x PF -1.
Rules for Access Categories (ACs) with indices i and j where 0 β€ i β€ j β€ 3: CWmin[i] β₯ CWmin[j], CWmax[i] β₯ CWmax[j], and AIFS[i] β₯ AIFS[j].
Priority Access Gateway Designation:
| Priority | AC | Designation |
|---|---|---|
| 0-2 | 0 | Best Effort |
| 3 | 1 | Video Probe |
| 4-5 | 2 | Video |
| 6-7 | 3 | Voice |
EDCF (Enhanced DCF)
EDCF is an enhanced version of DCF that supports bursting and uses different parameters for different TC/AC. It replaces DIFS with AIFS (Arbitration Interframe Space) , where AIFS β₯ DIFS. AIFS is shorter for audio and video traffic: Audio = Video < Data. For CWmin and CWmax: Audio < Video < Data. Different Persistence Factor (PF) values are used. For EDCF: AIFS β₯ DIFS, PF ranges from 1 to 16, and CWmin ranges from 0 to 255.
HCF (Hybrid CF)
HCF provides policing and deterministic channel access by controlling the channel through the HC (Hybrid Coordinator) . It can operate in both CFP and CP modes. The HC detects the channel as being idle for PIFS, which is shorter than DIFS, giving the HC high priority over EDCF. The HCF model can provide Guaranteed Services with a much higher probability than pure EDCF. A signaling protocol can be used to facilitate admission control and specify service rate requirements.
Mobile IP: Basics
Mobile IP was proposed by the IETF (Internet Engineering Task Force) , the standards development body for the Internet. Mobile IP allows a mobile host to move about without changing its permanent IP address. Each mobile host has a home agent on its home network. The mobile host establishes a care-of address when it's away from home.
A correspondent host is a host that wants to send packets to the mobile host. The correspondent host sends packets to the mobile host's permanent IP address. These packets are routed to the mobile host's home network. The home agent forwards IP packets for the mobile host to the current care-of address. The mobile host sends packets directly to the correspondent, using its permanent home IP as the source IP.
π Definition β Mobile IP: A protocol that allows mobile hosts to move between networks while maintaining the same permanent IP address.
Mobile IP: Care-of Addresses
Whenever a mobile host connects to a remote network, two choices exist for the care-of address:
- The address of a foreign agent on the remote network, where the foreign agent delivers packets forwarded from the home agent to the mobile host.
- A temporary, foreign IP address obtained through DHCP, where the home agent tunnels packets directly to the temporary IP address.
Regardless of the choice, the care-of address must be registered with the home agent.
π Example: A mobile node moves from its home network (192.168.1.0/24) to a foreign network (10.0.0.0/24). It acquires a care-of address of 10.0.0.5 (either from a foreign agent or via DHCP) and registers this with its home agent (192.168.1.1). Packets sent to the mobile's permanent address (192.168.1.100) are intercepted by the home agent and tunneled to 10.0.0.5.
Protocol
Mobile IP includes three capabilities:
-
Discovery: Mobile Agents send ICMP router advertisements with mobility agent advertisement extensions periodically, informing mobile nodes of their presence. The mobile node is responsible for the discovery process and may optionally request an advertisement from an agent.
-
Registration: The mobile node recognizes it is on a foreign network, acquires a Care-of-Address, and requests its home agent to forward its data packets to the foreign agent. Registration requires 4 steps:
- Mobile node sends registration request to the foreign agent
- Foreign agent relays request to the home agent
- Home agent accepts or denies and sends registration reply to the foreign agent
- Foreign agent relays reply to the mobile node
-
Tunneling: After registration, an IP tunnel is set up between the home agent and the care-of address. The home agent broadcasts a gratuitous ARP request causing all nodes in the subnet to update their ARP caches to map the mobile node's IP address to the home agent's link-level address. The home agent forwards packets to the foreign agent through the IP tunnel. In the foreign network, decapsulation is done by the foreign agent or the mobile node itself. The correspondent node assumes replies come from the home network and continues sending packets to the home agent.
IP-in-IP Tunneling
The packet to be forwarded is encapsulated in a new IP packet. In the new header:
- Destination = care-of address
- Source = address of home agent
- Protocol number = IP-in-IP
π Formula: IP-in-IP Encapsulation: Original IP packet β new IP header (dest: COA, src: HA) + original IP packet β tunneled to care-of address.
Triangle Routing (Mobile IPv4)
Triangle routing occurs when packets from the correspondent host go through the home agent, which then tunnels them to the mobile node's care-of address, while the mobile node sends replies directly back to the correspondent. This creates routing inefficiency, as packets must travel an extra hop through the home network.
π‘ Why this matters: Triangle routing introduces unnecessary delay and network load, especially when the correspondent and mobile node are geographically close while the home agent is far away.
Routing Inefficiency
The triangle routing problem creates significant inefficiency because:
- Packets travel longer paths than necessary
- The home agent becomes a bottleneck
- Latency increases for all communications
Route Optimizations
Possible Solution: The home agent sends the current care-of address to the correspondent host, which caches the care-of address. Future packets are tunneled directly to the care-of address. However, this introduces a cache consistency problem β the cached care-of address becomes stale when the mobile host moves. Additionally, there are potential security issues with providing the care-of address to the correspondent host.
Mobile IPv6 Roaming
Mobile IPv6 provides improved mobility support compared to MIPv4, with better route optimization built into the protocol design rather than as an add-on. IPv6's larger address space and built-in security features help address some of the limitations found in Mobile IPv4.
β Key Takeaways
Students must understand that legacy 802.11 DCF and PCF have fundamental QoS limitations β DCF only provides best-effort service with no guarantees, while PCF suffers from inefficient polling and unpredictable delays. The 802.11e standard addresses these issues through EDCF (prioritized channel access using AIFS, CWmin/max, and PF parameters per traffic category) and HCF (deterministic channel access via the Hybrid Coordinator using PIFS). For Mobile IP, the three core protocol capabilities are discovery (agent advertisements), registration (four-step process via foreign agent), and tunneling (IP-in-IP encapsulation). Triangle routing is a key inefficiency in Mobile IPv4, where packets go through the home agent even when the correspondent could send directly, and route optimization attempts to solve this but introduces cache consistency and security challenges.
π§ Quick Revision Questions
- What are the three QoS limitations of DCF and the three limitations of PCF in legacy 802.11?
- What are the four Access Categories in 802.11e and what are their relative priorities for AIFS, CWmin, and CWmax?
- Describe the formula for calculating CWnew[TC] in 802.11e and explain the role of the Persistence Factor.
- What are the four steps involved in the Mobile IP registration process, and which entities participate in each step?
- Explain triangle routing in Mobile IPv4 and identify two problems with the route optimization solution that caches care-of addresses at correspondent hosts.
π Lecture 31 β Wireless Mesh Networks
π Overview: This lecture introduces Wireless Mesh Networks (WMNs), a self-organizing, self-configuring network architecture that extends wireless coverage and integrates with existing networks. It covers WMN characteristics, compares WMNs with MANETs, explains the three primary architectures (infrastructure, client, hybrid), and discusses applications and critical performance factors.
ποΈ Topics Covered
The lecture begins with an introduction to WMN components and characteristics, then contrasts WMNs with MANETs across key dimensions like backbone structure and mobility. It details the three WMN architecturesβinfrastructure/backbone, client, and hybridβand explores applications including broadband home, community, and enterprise networking. The lecture concludes with critical factors influencing WMN performance such as radio techniques, scalability, and security.
π Lecture Summary
Introduction
Wireless Mesh Networks (WMNs) are comprised of mesh routers and mesh clients. Mesh clients not only work as hosts but also perform routing for multi-hop destinations. Mesh routers support bridge/gateway functionalities, enabling integration of WMNs with existing wireless networks such as cellular, WSN, WiMAX, etc. A WMN is dynamically self-organized and self-configured. Conventional nodes (e.g., PC, PDA, PocketPC, phones) equipped with a wireless NIC can connect directly to mesh routers; without a wireless NIC, an Ethernet connection is also possible. Thus, WMN will allow always-on-line anywhere anytime. It is gaining interest as a possible way for ISPs and can be deployed incrementally as needed. Deploying WMN is not difficult because most components/protocols are readily available (e.g., IEEE 802.11, WEP), however, scalability in existing protocols is a great concern.
Characteristics
WMNs have several key characteristics:
- Multi-hop: To extend the coverage range of wireless networks without sacrificing channel capacity and for non-line-of-sight.
- Support for ad hoc networking: Due to flexible architecture, easy deployment and configuration, fault tolerance, and mesh connectivity. This includes low up-front investment requirements.
- Mobility dependence on type of mesh nodes: Mesh routers usually have minimal mobility, while mesh clients can be stationary or mobile.
- Multiple type of network access: Both backhaul access to the internet and P2P communication are supported. Integration of WMN with other wireless networks allows end-users access to WMN.
- Dependence of power-consumption constraints on the type of mesh nodes.
- Compatibility and interoperability with existing wireless networks: A WMN based on IEEE 802.11 should support both mesh clients and Wi-Fi clients and be interoperable with other networks.
WMNs vs MANET
WMNs are considered ad hoc due to lack of infrastructure (AP/BS). Although ad hoc techniques like MANET are required, WMNs require more sophisticated algorithms and design principles. WMN diversifies the capabilities of ad hoc, making MANET a subset of WMNs. Differences include:
- Wireless Backbone: Mesh routers form a wireless backbone providing more coverage, connectivity, and robustness. In MANET, individual nodes are routers, making the network unreliable.
- Integration: WMN supports clients that use the same radio technology, accomplished through a host-routing function in the mesh router. Users of one network can enjoy services of another network.
- Dedicated routing and configuration: In MANET, each host performs routing and configurations; in WMN, this is done by mesh routers, decreasing the load on end-users.
- Multiple radios: WMNs use two radiosβone for routing/configuration between mesh routers, and a second for network access by end-users. In MANET, these functions are performed on the same channel, significantly degrading performance.
- Mobility: Hosts working as routers in MANET make it more challenging, whereas the mobility of mesh routers is very limited.
π Definition β MANET subset of WMNs: MANETs are a subset of WMNs because WMNs require more sophisticated algorithms and have diversified capabilities beyond simple ad hoc networking.
Architecture
Mesh Router
A mesh router supports routing functions for mesh networking in addition to conventional gateway/repeater functions. It is equipped with multiple interfaces built on either the same or different wireless access technologies. It achieves the same coverage as other wireless routers with less energy consumption through multi-hop routing. MAC protocols are enhanced with better scalability in a multi-hop mesh environment. Examples of mesh routers are based on embedded systems like PowerPC and Advanced Risc Machines (ARM).
Mesh Clients
Mesh clients also have necessary functions for routing in mesh networking; however, gateway or bridge functions do not exist. They usually have a single interface. Examples of mesh clients include laptops, PDAs, Wi-Fi IP Phones, and Wi-Fi RFID Readers.
Infrastructure/Backbone WMNs
This architecture is built using various wireless technologies in addition to 802.11. With gateway functions, a mesh router can connect to the internet. Infrastructure meshing allows integration of different networks. If a client has a different technology, it can connect through a BS, and the BS connects via Ethernet. This is the most common type. For example, community and neighborhood networks can be built where mesh routers are placed on rooftops, serving as access points for users inside houses and along roads. They typically have two types of radios: one long-range for the backbone and one short-range for end-users.
Clients WMNs
Client meshing provides a P2P network among client devices. No mesh router is required. Clients in this architecture require more functionalities for configuration and routing. It is typically formed using a single radio.
Hybrid WMNs
This is a combination of infrastructure and client meshing and is the most applicable/practical scenario.
Application scenario
Research and development in WMNs is motivated by several applications:
- Broadband home networking: A WLAN is not practical because an AP leaves dead zones, and multiple APs require a backbone network or access hub. Dead zones can be eliminated with multiple routers and adjusting their transmission power.
- Community and neighborhood networking: The common architecture uses cable or DSL through the internet, which has drawbacks: all traffic must flow through the internet, dead zones exist, services between end-users cannot be shared, and there is a single path for internet and neighborhood users.
- Enterprise networking: WMNs are scalable with enterprise growth.
- Other applications: Transportation systems (extending access into buses, trains, ferries; remote monitoring of in-vehicle security video), building automation, health and medical systems, and security surveillance systems.
Critical factors influencing network performance
- Radio techniques: Includes directional and smart antennas, MIMO systems, multi-radio chipsets, and more advanced techniques like reconfigurable and cognitive radios. These require revolutionary design changes in higher layers.
- Scalability: Multihop routing is common in WMN, which degrades performance. IEEE 802.11 MAC is not scalable, and throughput significantly reduces as the number of hops increases to 4 or higher.
- Mesh connectivity: Network self-organization and topology control algorithms are needed.
- Broadband and QoS: Most applications of WMN are broadband services with various QoS requirements, which differs from ad hoc networks.
- Compatibility and inter-operability: Network access must be provided to both conventional and mesh clients.
- Security: There is no centralized control.
- Ease of use.
β Key Takeaways
WMNs use mesh routers as a dedicated wireless backbone to enhance coverage, robustness, and reliability, unlike the unreliable node-based routing in MANETs. The architecture can be infrastructure-based (most common, using two radios for backbone and users), client-based (P2P, no routers), or hybrid. WMNs provide significant advantages over basic WLANs by eliminating dead zones and supporting incremental deployment, but they face critical challenges including poor scalability of IEEE 802.11 MAC (throughput drops significantly beyond 4 hops), the need for advanced radio techniques, and security concerns due to no centralized control.
π§ Quick Revision Questions
- What are the two main components of a Wireless Mesh Network, and what are their key roles?
- How does mobility differ between mesh routers and mesh clients in a WMN?
- List three key differences between WMNs and MANETs regarding backbone, routing, and radios.
- What are the three types of WMN architectures, and which is the most practical?
- Name three critical factors that influence the performance of a Wireless Mesh Network.
π Lecture 32 β Wireless Mesh Networks Part II
π Overview: This lecture continues the study of Wireless Mesh Networks (WMNs), focusing on the critical challenges and solutions at the MAC and Network layers. It examines scalability issues, single versus multi-channel MAC protocols, routing requirements and optimization criteria, cross-layer design, fairness problems, and existing WMN standards, providing a comprehensive view of how these networks operate efficiently.
ποΈ Topics Covered
The lecture covers MAC layer challenges including scalability, single channel and multi-channel approaches, and open research issues. It then delves into the network layer, discussing routing protocols, a wish list for ideal routing, optimization criteria, fairness, and cross-layer design. Finally, it addresses QoS support required at every layer and presents WMN standards like IEEE 802.16a and 802.11s.
π Lecture Summary
MAC Layer
The MAC layer for WMNs must handle more than one-hop communication in a distributed and cooperative manner, supporting multipoint-to-multipoint communication. Network self-organization is needed, and mobility affects performance. Scalability can be addressed by either enhancing single-channel approaches to increase end-to-end throughput or allowing multiple channels per node. π Definition β Scalability: The ability of a MAC protocol to maintain performance as the number of nodes in the network increases.
Basic MAC techniques are divided into Scheduled (e.g., fix scheduled TDMA, polling) which are often impractical due to lack of central coordination and time synchronization, and Random Access (e.g., CSMA, RTS/CTS). Channels can be implemented via FDMA, CDMA (where code assignment is an issue), SDMA with directional antennas, or combinations thereof.
Single channel MAC
Approaches to improve single-channel MAC include: modifying parameters of CSMA/CA based MACs like contention window size and backoff procedures for multi-hop; cross-layer design with advanced physical layers like directional antenna and power control; and innovative new solutions like TDMA or CDMA based MACs. Compatibility, cost, and complexity are key design factors. Contention-based approaches are not scalable, and throughput degrades with increased contention.
MAC β Multichannel
Multi-channel approaches increase network capacity by using multiple channels of bandwidth B. A Multi-Channel Coordination Layer (MCCL) can be introduced between the MAC and Network layers to work within 802.11 constraints and increase capacity.
Three types of multi-channel MAC exist:
- Multi-channel single transceiver: One active channel at a time; coordination is required as different nodes may operate on different channels at different times.
- Multi-channel multi-transceiver: A single radio has multiple parallel RF chips and baseband modules to support several simultaneous channels, with a single MAC coordinating operations.
- Multi-radios MAC: A node has multiple radios, each with its own MAC and physical layer; a virtual MAC protocol like Multi-radio Unification Protocol (MUP) coordinates communication among all.
MMAC functions include: maintaining a data structure of all channels (classified by allocation status), negotiating channels during the Ad hoc Traffic Indication Message (ATIM) window on a predefined common channel, and selecting the channel with the lowest count of source-destination pairs.
Multi-radio Unification Protocol (MUP) functions include: discovering neighbors and classifying them as MUP-enabled or legacy; selecting a NIC based on one-hop round trip time (RTT) measurements (choosing the shortest RTT); utilizing the selected NIC for a random period (10-20 seconds); and switching channels after that period if another NIC shows a certain quality improvement.
Open research issues remain: scalability in multi-hop ad hoc networks is unsolved; CSMA/CA solves only partial problems; a distributed TDMA or CDMA MAC may be a solution; mesh routers and clients have different characteristics (mobility, power); a single solution may not work for both; bridging functions are needed for routers integrating multiple technologies; and MAC must be developed with QoS metrics for broadband multimedia communication.
Network Layer
WMNs are tightly coupled with the internet, but their routing differs from IP and cellular routing.
Routing finds and maintains routes for data flows. The entire performance of the WMN depends on the routing protocol, which may be the main product of a mesh company.
π Routing β Wish List: An ideal routing protocol should include scalability (low overhead, especially in mobile WMNs), fast route discovery and rediscovery (essential for reliability), mobile user support (seamless handover), flexibility (work with/without gateways and different topologies), QoS support (satisfying specified criteria), and multicast support (important for applications like emergency response).
Existing Routing Protocols include Internet routing protocols (e.g., OSPF, BGP, RIPv2) which are well-known but assume seldom link changes and are unsuitable without modifications. Ad-hoc routing protocols (e.g., DSR, AODV, OLSR, CBR, TORA) are designed for high rates of link changes and perform well on WMNs, but may be further optimized.
Routing - Optimization Criteria
Routes can be optimized based on several criteria, individually or in combination: π Formula: Minimum Hops β shortest path π Formula: Minimum Delays β path with least latency π Formula: Maximum Data Rates β path with highest bandwidth π Formula: Minimum Error Rates β path with fewest errors π Formula: Maximum Route Stability β path that is likely to last longest π Formula: Power Consumption β path that uses least energy Other criteria include use of multiple routes to the same gateway or use of multiple gateways.
Routing β Cross-Layer Design
Cross-layer design integrates routing with other layers:
- Routing β Physical: Link quality feedback helps select stable, high-bandwidth, low-error routes. Fading signal strength can signal a link about to fail, enabling preemptive route requests. Cross-layer design is essential for smart antennas.
- Routing β MAC: Feedback on link loads can avoid congested links, enabling load balancing. Channel assignment and routing depend on each other. MAC detection of new neighbors and failed routes improves routing performance.
- Routing β Transport: Choosing low-error-rate routes improves TCP throughput, especially with multiple routes. Freezing TCP when a route fails is helpful.
- Routing β Application: Especially important for satisfying QoS constraints.
Network Layer - Fairness
Fairness provides an equal share of resources to all participants and is a special case of priority-based QoS.
- Horizontal fairness (between nodes 1 and 2): Ensured by the MAC layer.
- Vertical fairness (between nodes 3 and 4, where node 3 is further from the gateway): The MAC layer is no longer sufficient because nodes closer to the gateway have an inherent advantage. π‘ Why this matters: In a multi-hop WMN, nodes closer to the gateway can starve nodes farther away, as local MAC fairness does not account for multi-hop traffic flows.
Fairness Problem: Nodes closer to the gateway (e.g., node 2) get excessive throughput, while node 3, which is one hop further, gets very little throughput, even though the MAC layer appears fair.
QoS Support required at every layer
QoS must be supported at every layer:
- Physical Layer: Requires robust modulation and link adaptation.
- MAC Layer: Must offer priorities and guarantees (bandwidth, delay).
- Network Layer: Must select "good" routes, offer priorities, and reserve resources for guarantees.
- Transport: Attempts end-to-end recovery when possible.
- Application: Negotiates end-to-end and with lower layers, adapting to changes in QoS.
WMNs Standards
Key standards include:
- WPAN: Bluetooth, Zigbee
- WiFi: 802.11a, b, g, n
- WiMAX: 802.16
IEEE 802.16a WMAN Mesh: Defines a "mesh mode" in addition to point-to-multipoint (PMP) mode. It operates in licensed and unlicensed frequencies of 2β11 GHz, allows non-line-of-sight (NLOS) communications, and spans up to a 50 km range, supporting multihop communications.
802.11s WLAN Mesh: Extends the MAC layer to a wireless Distribution System (DS) to support broadcast/multicast, adds multi-hop capability to 802.11g/a/b, and features auto-configuration on power up, multi-channel multi-radio operation, topology discovery, MAC path selection protocol, and modified forwarding for QoS and mesh control.
802.11s MCF Sublayer: The Mesh Coordination Function (MCF) sublayer is part of the 802.11s standard.
β Key Takeaways
The critical points from this lecture are that WMN MAC protocols face scalability issues, which can be addressed through single-channel enhancements or multi-channel approaches like MMAC and MUP. Network layer routing must be optimized not just for hops but for cross-layer metrics like link quality and load, and must ensure both horizontal and vertical fairness, as nodes closer to gateways can starve others. Finally, QoS must be supported at every layer from physical to application, and protocols like 802.16a and 802.11s are key standards for implementing WMNs.
π§ Quick Revision Questions
- What are the two main approaches to addressing scalability in the MAC layer for WMNs?
- Describe the three different types of multi-channel MAC architectures (single transceiver, multi-transceiver, multi-radio) and a key challenge for each.
- What is the "fairness problem" in WMNs and why is the MAC layer alone insufficient to solve it?
- List the five layers at which QoS support is required in a WMN and give one example of a QoS mechanism for each.
- What is the role of the Multi-radio Unification Protocol (MUP) and how does it select a Network Interface Card (NIC)?
π Lecture 33 β TCP over Wireless Networks
π Overview: This lecture addresses the fundamental challenges of running TCP over wireless networks, where packet loss is often due to transmission errors or mobility rather than congestion. It explores various TCP variants designed to improve performance in wireless environments and categorizes the primary approaches for adapting TCP to heterogeneous wireless networks.
ποΈ Topics Covered
The lecture covers the motivation for TCP modifications in wireless networks, details of TCP variants including Slow Start and Fast Retransmit/Recovery (TCP Reno), the specific issues in heterogeneous wireless networks, and four major categories of TCP schemes for wireless: Pure Link-level Approaches, Soft-state Transport Layer Caching Approaches (including Snooping TCP), Soft-state Cross Layer Signaling Approaches (including ECN, EBSN, ELN), and Hard-state Transport Layer Approaches (including Indirect TCP, Mobile TCP, Wireless TCP, ATP, and ATCP).
π Lecture Summary
Motivation
Traditional TCP was designed for fixed, wired networks with characteristics including window-based control, slow-start, loss-based congestion indication, and dependence on ACKs. Wireless networks present fundamentally different conditions: lack of infrastructure in ad hoc networks, mobility, shared channels, and limited bandwidth. Packet loss in fixed networks is typically due to overload and detected via retransmission timeout (RTO), three duplicate ACKs, or ICMP source quench messages. Routers discard packets when buffers are full, and TCP recognizes congestion only indirectly through missing acknowledgements. Retransmissions during congestion worsen the problem.
π‘ Why this matters: Directly applying wired TCP to wireless networks causes severe performance degradation because TCP misinterprets wireless transmission errors as congestion signals.
π Definition β TCP congestion control: A mechanism where the sender detects packet loss (assumed due to congestion) and reacts by reducing its transmission rate.
TCP Slow Start
The sender calculates a congestion window for a receiver. The process begins with a congestion window size equal to one segment, then an exponential increase of the congestion window up to the congestion threshold, followed by a linear increase. A missing acknowledgement causes reduction of the congestion threshold to one half of the current congestion window. The congestion window then starts again with one segment.
π Formula (implicit): Congestion window (cwnd) increases exponentially until threshold, then linearly; on loss, threshold = cwnd/2, cwnd = 1.
π Example: If cwnd starts at 1 segment and threshold is 16, the sender sends 1 segment, then 2, then 4, then 8, then 16 (exponential increase). After reaching 16, it increases linearly (17, 18, etc.). If a loss occurs at cwnd=20, threshold becomes 10 and cwnd resets to 1.
TCP Fast Retransmit/Recovery (TCP Reno)
TCP sends an acknowledgement only after receiving a packet. If a sender receives several acknowledgements for the same packet, this indicates a gap in received packets at the receiver. The receiver has received all packets up to the gap and is still receiving packets, so packet loss is not due to congestion. When packet loss occurs, congestion window size is reduced: due to timeout, cwnd = 1 and enter slow start; due to duplicate ACKs, cwnd = cwnd/2 + 3 Γ segment_size. Congestion window size is increased when data is successfully acknowledged.
π Definition β Fast Retransmit/Recovery: A TCP mechanism that detects packet loss via duplicate ACKs (not timeout) and reduces the congestion window less drastically than slow start (halving rather than resetting to 1).
Issues in Heterogeneous Wireless Networks
Key challenges include:
- Bit Error Rate (BER): 10β»Β³ or worse possible upon change in wireless environment
- Bandwidth: Very less compared to wired networks; TCP underestimates bandwidth in wireless networks
- Round Trip Time (RTT): Wireless media exhibits longer latencies due to long distances or NLOS path; large variation in RTT
- Mobility: Addition of mobile devices introduces huge amounts of indeterminate delay
- Power consumption
Influences of BER/mobility on TCP: TCP assumes congestion if packets are dropped, which is typically wrong in wireless networks where packet loss is often due to transmission errors. Mobility itself can cause packet loss when a mobile node roams between access points with packets still in transit. The performance of unchanged TCP degrades severely. However, TCP cannot be changed fundamentally due to the large base of installation in the fixed network.
π‘ Why this matters: Distinguishing between congestion loss, error loss, delay beyond retransmission timer threshold, and out-of-order delivery beyond the three DUPAK threshold is critical for wireless TCP performance.
TCP Schemes for Wireless β Classification
The various approaches revolve around distinguishing between:
- Congestion loss
- Error loss
- Delay beyond the retransmission timer threshold
- Out-of-order delivery beyond the three DUPAK threshold
Classification categories: Pure Link-level Approaches, Soft-state Transport Layer Caching Approaches, Soft-state Cross Layer Signaling Approaches, and Hard-state Transport Layer Approaches.
Pure Link-level Approaches
These approaches aim at hiding the unwanted characteristics of the wireless links from the higher layers. Reliable link-level protocols are implemented on the wireless link which perform local retransmissions independent of the higher-level protocols. These protocols employ techniques such as forward error correction (FEC) for error control and automatic repeat request (ARQ) for retransmission of lost packets. The timeout value for local retransmissions is of major concern β interaction between link-level retransmission timeouts and transport-level timeouts can lead to degraded performance if not properly selected.
π Definition β Pure Link-level Approaches: Methods that hide wireless link characteristics from higher layers by implementing reliable local retransmissions at the link level.
Soft-state Transport Layer Caching Approaches β Snooping TCP
Snoop involves modification of the network layer (IP) software at the base station (BS) by adding a module called snoop. It provides transparent extension of TCP within the BS/FA with buffering of packets sent to the mobile host. Lost packets on the wireless link (both directions) will be retransmitted immediately by the mobile host or foreign agent (local retransmission). The foreign agent "snoops" the packet flow and recognizes acknowledgements in both directions, filtering ACKs.
Data transfer to the mobile host: FA buffers data until it receives ACK of the MH; FA detects packet loss via duplicated ACKs or time-out with fast retransmission possible, transparent for the fixed network.
Data transfer from the mobile host: FA detects packet loss on the wireless link via sequence numbers; FA answers directly with a NACK to the MH; MH can retransmit data with very short delay.
Advantages: Maintains end-to-end semantics, no change to correspondent node, no major state transfer during handover.
Problems: Snooping TCP does not isolate the wireless link well; snooping might be useless depending on encryption schemes.
π Definition β Snooping TCP: A soft-state approach where a base station module buffers packets and performs local retransmissions transparently to the fixed network.
Soft-state Cross Layer Signaling Approaches
- Explicit Congestion Notification (ECN): Extension to Random Early Detection (RED); marks a packet instead of dropping when the average queue size is between minth and maxth. Upon receipt of a congestion-marked packet, the TCP receiver informs the sender about incipient congestion, triggering the congestion avoidance algorithm.
- Explicit Bad State Notification (EBSN): Proposes a mechanism to update the TCP timer at the source to prevent the source from decreasing its congestion window. EBSNs are sent to the source after every unsuccessful attempt by the base station to transmit packets over the wireless link. EBSN cancels previous timeouts and puts new timeouts in place based on existing estimates of RTT and variance.
- Explicit Loss Notification (ELN): Adds an ELN option to TCP ACKs. When a packet is dropped on wireless networks, future cumulative acknowledgements corresponding to the lost packet are marked to identify that a non-congestion-related loss has occurred.
π Definition β Explicit Congestion Notification (ECN): A signaling mechanism where routers mark packets instead of dropping them to indicate incipient congestion.
Hard-state Transport Layer Approaches β Indirect TCP (I-TCP)
Indirect TCP or I-TCP segments the connection. No changes to the TCP protocol for hosts connected to the wired Internet; optimized TCP protocol for mobile hosts. The splitting of the TCP connection occurs at the foreign agent into 2 TCP connections, with no real end-to-end connection any longer. Hosts in the fixed part of the net do not notice the characteristics of the wireless part.
Advantages: No changes in the fixed network necessary; all current optimizations to TCP still work; transmission errors on the wireless link do not propagate into the fixed network; simple to control; mobile TCP is used only for one hop between foreign agent and mobile host; very fast retransmission possible.
Disadvantages: Loss of end-to-end semantics β an acknowledgement to a sender no longer means a receiver really got a packet; foreign agents might crash; higher latency possible due to buffering and forwarding to a new foreign agent.
Wireless TCP (WTCP)
Preserves the end-to-end semantics. This protocol tries to distinguish random losses from congestion losses by measuring packet inter-arrival time with packet inter-departure time. WTCP uses rate-based rather than window-based transmission control, shaping its data traffic and never allowing a burst of packet transmissions. It is fair when competing connections have different round-trip times.
Mobile TCP (M-TCP)
Special handling of lengthy and/or frequent disconnections with low BER links. M-TCP splits as I-TCP does: unmodified TCP from fixed network to supervisory host (SH), optimized TCP from SH to MH. The supervisory host monitors all packets, no caching, no retransmission. If disconnection is detected, it sets the sender window size to 0, causing the sender to automatically go into persistent mode. Old or new SH reopen the window.
Advantages: Maintains semantics, supports disconnection, no buffer forwarding. Disadvantages: Loss on wireless link propagated into fixed network; adapted TCP on wireless link.
Ad Hoc Transport Protocol (ATP)
Features include: layer coordination using feedback from network nodes for congestion detection, avoidance, and control; rate-based transmissions to avoid the impact of bursty traffic; decoupling of congestion control and reliability (congestion control uses network feedback, reliability uses receiver feedback and selective ACK); assisted congestion control adapting sending rate based on feedback from intermediate nodes; TCP friendliness and fairness achieved through feedback, though fairness remains an issue.
ATCP Approach
ATCP utilizes network layer feedback from intermediate nodes to take appropriate actions. Network feedback includes: ICMP (Destination Unreachable message indicates route disruption) and ECN (indicates network congestion). With ECN enabled, timeout and 3 duplicate ACKs are assumed to no longer be due to congestion.
Transport Layer Challenges
New transport layer protocols need to be developed that avoid the shortcomings of TCP while being compatible with it. Transport layer protocols for supporting real-time traffic in wireless meshes are desirable. Integration of transport layer with other layers is needed, as is inferring and reacting with respect to observations at other layers. The impact of mobility on transport layer must be addressed.
β Key Takeaways
The single most critical concept is that standard TCP assumes all packet loss is due to congestion, which is fundamentally wrong in wireless networks where loss comes from transmission errors and mobility. Students must understand the four categories of wireless TCP schemes: Pure Link-level (hiding wireless characteristics), Soft-state Caching (snooping approaches maintaining end-to-end semantics), Soft-state Cross Layer Signaling (ECN, EBSN, ELN for explicit notification), and Hard-state (I-TCP splitting the connection, sacrificing end-to-end semantics). The key trade-off is between maintaining end-to-end semantics and effectively isolating the wireless link. Snooping TCP and I-TCP represent opposite ends of this spectrum. Finally, the lecture emphasizes that any solution must remain compatible with the existing TCP installed base in the fixed Internet.
π§ Quick Revision Questions
- Why does standard TCP perform poorly in wireless networks, and what specific characteristics of wireless networks cause this degradation?
- What is the difference between TCP Slow Start and Fast Retransmit/Recovery in terms of congestion window adjustment after packet loss?
- How does Snooping TCP maintain end-to-end semantics while performing local retransmissions at the base station?
- What are the three explicit notification mechanisms (ECN, EBSN, ELN) and how does each help distinguish congestion loss from error loss?
- What is the fundamental trade-off between Indirect TCP (I-TCP) and approaches that maintain end-to-end semantics?