Introduction to Wireless Networking CS 490WN/ECE 401WN Winter 2007 Lecture 4: Wireless LANs and IEEE 802.11 Part II This lecture continues the study of wireless LANs by looking at IEEE 802.11. I. 802.11 Architecture and Services 802.11 Working Group Chapter 14 Wi-Fi and the IEEE 802.11 Wireless LAN Standard (continued) Started in 1990 To develop MAC protocol and physical medium specifications. And use existing 802 LLC functions. Initial interest was to use unlicensed frequencies. Called the ISM (Industrial, Scientific, and Medical) bands in U.S. The 802.11 Working Group has an ever-expanding list of standards. Lecture 4, Page 1 of 20
This is that latest list of active groups from the 802.11 web site (http://www.ieee802.org/11/), since some in the above table are no longer active. 802.11k, Radio Resource Measurement of Wireless LANs - Radio Resource Measurement enhancements to provide mechanisms to higher layers for radio and network measurements. 802.11m, Maintenance PAR 802.11n, High Throughput 802.11p, Wireless Access in the Vehicular Environment - enhancements to 802.11 required to support Intelligent Transportation Systems (ITS) applications 802.11r, Fast Roaming Fast Handoff 802.11s, Mesh Networking 802.11T, Wireless Performance Prediction 802.11u, Wireless Interworking With External Networks 802.11v, Wireless Network Management 802.11w, Protected Management Frames 802.11y, Contention Based Protocol Study Group Here is a table of 802.11 terminology. Lecture 4, Page 2 of 20
Wi-Fi Alliance The first 802.11 standard to gain industry acceptance was 802.11b. 2.4 GHz, up to 11 Mbps. There was concern whether products would successfully interoperate Linksys Access Point with a Cisco PCMCIA card? Wireless Ethernet Compatibility Alliance (WECA) formed in 1990. Renamed the Wi-Fi (Wireless Fidelity) Alliance. Certifies interoperability for 802.11b products. Certified products are called Wi-Fi. - Over 100 products certified. Certification has been extended to 802.11g products. - 802.11g is at 2.4 GHz, up to 54 Mbps. The Wi-Fi Alliance is concerned with markets for WLANs in enterprises, homes, and public hot spots. The 802.11 Architecture See Figure 14.4 Lecture 4, Page 3 of 20
The smallest building block of a WLAN is a. Stations executing the same MAC protocol. Stations competing for access to the same shared wireless medium. Two BSSs can overlap geographically. - A single station can participate in more than one BSS. - Using different frequency bands. BSSs connect through a Distribution System (DS). Can be a switch, a wired network, or a wireless network. The is the bridge and relay point. Stations do not communicate directly with each other. - But through the AP. An AP is part of a station. - STA1 and STA5 above. If there is no connection to other BSSs, the BSS is called an. Stations can communicate directly. No AP is necessary. An Extended Service Set (ESS) consists of two or more BSSs connected by a distribution system. The ESS appears as a single logical LAN to LLC. Lecture 4, Page 4 of 20
IEEE 802.11 Services Nine services are provided to give functionality equivalent to wired LANs. Two ways the services are categorized. 1. Provided by a station or by provided by the distribution system. 2. LAN access and security versus delivery of MAC packets (called MAC Service Data Units MSDUs). MSDU Delivery Service MSDUs are the blocks of data passed down from the MAC user (like the LLC layer) to the MAC layer. This service executes the delivery. And if the MSDU is too large, it may be broken into smaller frames. - This process is called. Services for distribution of messages within a DS Distribution Between BSSs. - APs are aware of each other and work together. - Specifics of the DS are beyond the scope of the 802.11 standard. Integration Between an 802.11 LAN and another 802.x LAN. - Takes care of translation and media conversion logic. Lecture 4, Page 5 of 20
Services for Association To transfer MSDUs, stations must be known to the WLAN. - To know where a destination station is located. A station must be associated. - Before it can deliver or accept data. Types of mobility to support. - No transition only movement within the range of a BSS. - BSS transition to another BSS in the same ESS. - Addressing capabilities must recognize the new location. - Hopefully with a fast, seamless transition (no disruption of service from the viewpoint of the users, on the order of 10 s of milliseconds). - ESS transition to another ESS. - Likely will cause a service disruption in this case. Services - Association - Associate with an AP. - APs share information with other APs. - Reassociation - Transfer an association to another AP. - Disassociation - Hopefully tell AP when leaving. - MAC management facility also protects itself against stations that disappear without disassociating. Services for Access and Privacy WLANs cannot rely on a physical connection for security. - In wired LANs, a station must be physically connected to be valid on the network. - WLANs are open to anyone within radio range. Services - Authentication - Establishes the identity of stations to each other. - Several authentication schemes are supported. - And also allows for expansion of the functionality. - The standard does not mandate any particular authentication scheme. - But whatever is used must be mutually acceptable to stations and APs. - Deauthentication Lecture 4, Page 6 of 20
- Privacy - Prevents contents of messages to be read by unintended recipients. - Encryption can optionally be used. II. 802.11 Medium Access Control Reliable Data Delivery Two main functions 1. Reliable Data Delivery 2. Sharing of the Wireless Medium (Medium Access Control) Reliable Data Delivery The wireless medium is extremely unreliable. For what reasons might frames (MAC layer packets) be lost? Noise Interference from other transmitters Propagation loss (weak signals) Signal reflections and fading We will discuss these in more detail in later lectures. Lecture 4, Page 7 of 20
Higher layer protocols like TCP will retransmit packets if they are not received properly at their ultimate destination. A path to a destination may cover many networks and wired or wireless links. But why might we want to retransmit just over the single wireless link? Timers waiting to recover lost packets end-to-end might take several seconds. MAC layer can retransmit more quickly. Simple Acknowledgements Destination stations send acknowledgement (ACK) frames back to their source stations. Both stations are on the 802.11 network. This is an atomic operation. Atomic an operation that is an indivisible unit. The data frame must be received, then the ACK must be sent back and received. - Before any other frames are sent. - Before anyone else can send anything over the medium. If the source does not receive the ACK within a certain period of time, it retransmits. Based on the size of the data frame and expected size of the ACK frame, the source knows when the ACK should have been received. This is a two-frame exchange for reliability. If we need more reliability, a four frame exchange can be used, as discussed below. Lecture 4, Page 8 of 20
RTS/CTS Additional Reliability Problem: All users may not be able to hear each other. X Problem (a) above: Hidden node problem C is sending to B. A cannot hear C and thinks it could also transmit to B. A s and C s packets will collide at B. Problem (b): Exposed node problem. B wishes to send to C. A is transmitting to station X. If B listens, it will think the radio channel is busy. So it will falsely conclude it cannot send to C. But C would hear no interference if B sent a packet to it. - C would not hear the one from A, since C is out of range from A. So, B could have transmitted but will not. Lecture 4, Page 9 of 20
Note that this scenario does not apply to 802.11. In 802.11, nodes B and C could still not communicate. Why? Because C would need to send an ACK to B, which would collide with the packet from A. Nonetheless, the exposed node problem was discussed here since it is commonly discussed. Solution: RTS and CTS Potential senders send a Request to Send (RTS). - Tells how long of a message it wishes to send. Potential receiver sends a Clear to Send (CTS) in response. - Also tells how long of a message will be sent. Assumption: If I hear something from Y, I am in Y s range and Y is in mine. Becomes a four-frame exchange RTS, CTS, Data, ACK. How does this RTS/CTS approach solve the hidden node problem? In (a), A will hear the CTS sent from B and will assume that if A sends a packet to B it will collide. Lecture 4, Page 10 of 20
How does this solve the exposed node problem? In (b), B will hear the RTS sent from A to X. But B will not hear the CTS from X, so B can assume that whatever it sends will not affect X. Notes: This assumes all stations have the same range. Collisions still might occur between RTS messages. Summary: Hear a CTS, don t send, since you are within range of the receiver of another communication. Only hear an RTS: - For one-way communication, assume okay, since the receiver of the other communication is out-of-range. - For two-way, don t send. The ACK will cause a problem even if the forward communication would not. For 802.11: The CTS and RTS clear out areas around active senders and receivers for no one to try to also communicate. This approach is called CSMA/CA Carrier Sense Multiple Access (CSMA) - Stations listen to the channel Collision Avoidance (CA) - RTS/CTS are used to prevent collisions of data packets Lecture 4, Page 11 of 20
Most implementations of 802.11 do not use the RTS/CTS procedure. Why? Less time delays, better throughput. Communication is usually with the AP. Exposed node is not an opportunity. Stations are usually within range. III. 802.11 Medium Access Control Random Access Two ways to share a medium 1. Distributed Access all nodes participate in the decisions to transmit packets. Good for ad hoc networks. May also be attractive for WLANs that have primarily bursty traffic. 2. Centralized Access regulation by a centralized decision maker. This is natural when a group of stations are connected together through a base station. Especially useful if some traffic is time sensitive or high priority. Examples: Voice over IP, video. Lecture 4, Page 12 of 20
So which did 802.11 implement? And the decision is... Result is DFWMAC (Distributed Function Wireless MAC). Distributed access mechanism with centralized control built on top. Distributed Coordination Function (DCF) Lower sublayer of the MAC layer. Uses a contention algorithm. Ordinary asynchronous traffic uses DCF. Point Coordination Function (PCF) Centralized MAC algorithm. Provides contention-free service. Exploits features of DCF also. Lecture 4, Page 13 of 20
Distributed Coordination Function (DCF) Makes use of simple CSMA If a station has a frame to transmit. Listen to the medium. - If the medium is idle, station may transmit. - Otherwise wait until the current transmission is complete. Once sending, collision detection is not used. - Unlike wired Ethernet which uses collision detection (CSMA/CD). - In CD, one would stop the communication once a collision is detected. - Because wireless signals are hard to distinguish from noise. - Hard to detect incoming colliding signals and separate them from a station s own transmission. - So, 802.11 just relies upon ACK s to indicate successful transmission. - Which means the station completes the entire transmission and completes the entire waiting period for an ACK. - This is wasteful, but unavoidable. DCF makes use of a type of time delays called an There are several types of IFSs. - More on that a little later in the lecture. Lecture 4, Page 14 of 20
Figure 14.6 gives the rules for CSMA access. Special notes Exponential backoff - More details are given here than in the book, see Giuseppe Bianchi, Performance Analysis of the IEEE 802.11 Distributed Coordination Function, IEEE Journal On Selected Areas In Communications, Vol. 18, No. 3, March 2000. - Wait a random time while the medium is idle, here s how. - Initialize a counter randomly selected in [0, CW-1]. - CW is the Congestion Window and starts at a given initial value (usually 16 or 32). - During every slot time the channel is sensed idle, decrease by the counter by 1. Lecture 4, Page 15 of 20
- If a channel is sensed busy, stop decrementing the counter until the channel is idle again, which might be considered a freeze time. - Once the counter reaches zero, transmit the frame. - If a collision occurs, start over with CW new =2* CW old. - A collision is assumed if an ACK is not received. - Keep doubling CW at each collision, up to a maximum value (usually 1024). - Called Exponential Backoff because in general an exponential increase is involved: CW i = 2 i * CW initial. - Also is more accurately called because each time the window is increased by a factor of 2. - Research is considering other values than 2. - Research is also considering different beginning CW and max CW. - What does exponential backoff accomplish? Stations wait long enough to give everyone a fair share of the wireless medium. - At the expense of what? At the expense of longer delays. - We at UMKC are studying several issues. - How stations share the medium under this scheme. - What if stations misbehave and do not follow standard behavior. Like what would be an example of misbehavior? Not expanding the CW, even keeping the CW=1. Lecture 4, Page 16 of 20
- How can stations have different control parameters (min CW, max CW, etc.) to fit their quality requirements, yet not hurt the performance of others. We call this MAC friendliness. There is not always a backoff on the first attempt - If the channel is idle for an IFS, then transmit immediately. - Why is this a good idea? If the channel is lightly loaded, then there is no unnecessary waiting for backoff. Functional priority-based access is provided using three types of IFSs SIFS Short IFS - Used for immediate response actions. PIFS Longer IFS used for the PCF. DIFS Longest IFS used for the DCF. Lecture 4, Page 17 of 20
Different types of messages get earlier access to the channel, based which type of IFS they use. The following use SIFS. - ACKs Recipients send back ACKs after only an SIFS. - Keeps the send/ack combination atomic. - Multiple MAC frames If an LLC protocol data unit is fragmented into multiple MAC frames, fragments (and all their ACKs) are sent before other packets using SIFS. - Keeps the whole LLC frame atomic. - CTS Sent after an SIFS in response to an RTS. - Poll response Associated with PCF as discussed later. PCF has priority over DCF because the PIFS is shorter than the DIFS. EIFS - There is an additional Interframe Space not shown in the above figure, called the Extended Interframe Space (EIFS). - This is longer than the DIFS (and all other IFSs). - It is to be used after a frame has been indicated by the physical layer to have been received as bad. Point Coordination Function The point coordinator issues polling messages. A polled station gets to send when it receives a poll, after only SIFS. Stations are polled in a round robin fashion, with PIFS between polls. It is unclear from the book if this could be weighted round robin. - In weighted round robin, some stations may get more than one chance to send during a given cycle. PCF uses polling to control the performance of time-sensitive traffic. - Like audio, video, etc. As long as the point coordinator sends polls, DCF is shut out. Because PCF messages will always be sent with PIFS spacing before DCF messages get a chance with DIFS spacing. But PCF operations will only occupy part of the time. Based on a time called a. - See part (b) of the Figure 14.7 above. Lecture 4, Page 18 of 20
PCF will occupy the first part of the superframe. - Then DCF operations will be allowed while the point coordinator idles. If the medium is busy at the end of a superframe, the next PCF period will be delayed. PCF is not used very often in 802.11 networks. Even though APs are used, they usually don t use polling. Little WLAN traffic (so far) needs such time-sensitive performance. - Especially if the bit rate of the WLAN is high enough that performance is good enough anyway. The IEEE 802.11e standard was developed as another way to provide better quality of service without using PCF. - Details on this in the next lecture. Medium Access Control Logic Figure 14.6 above shows that the exponential backoff is not used if a station finds the medium idle when it first wishes to send a packet. Let s confirm this form the 802.11 standard. The 802.11 standards can be obtained from the UMKC campus (or using the UMKC VPN) from http://ieeexplore.ieee.org under Standards. - Virtually all papers ever published by the IEEE are also available there. It is called Basic Access. Here is some of what the standard says. 9.2.5.1 Basic access Basic access refers to the core mechanism a STA uses to determine whether it may transmit. In general, a STA may transmit a pending MPDU when it is operating under the DCF access method, either in the absence of a PC, or in the CP of the PCF access method, when the STA determines that the medium is idle for greater than or equal to a DIFS period, or an EIFS period if the immediately preceding medium-busy event was caused by detection of a frame that was not received at this STA with a correct MAC FCS value. If, under these conditions, the medium is determined by the carrier-sense mechanism to be busy when a STA desires to initiate the initial frame of one of the frame exchanges described in 9.7, exclusive of the CF period, the random backoff algorithm described in 9.2.5.2 shall be followed. There are conditions, specified in 9.2.5.2 and 9.2.5.5, where the random backoff algorithm shall be followed even for the first attempt to initiate a frame exchange sequence. Lecture 4, Page 19 of 20
Next lecture: Last part (Part 3) of 802.11, related to the MAC frame, physical layers, other 802.11 standards, and 802.11 security. Lecture 4, Page 20 of 20