1. THROUGHPUT ANALYSIS OF IEEE 802.11 DCF BASIC IN PRESENCE OF HIDDEN STATIONS Shahriar Rahman Stanford Electrical Engineering http://ee.stanford.edu

2. Outline of Talk  802.11 DCF Protocol Overview  Problem with DCF Basic Access  Modeling Hidden Stations  DCF Throughput Models  Simulation Results  Discussions & Conclusion  Future Work  Q&A

3. IEEE 802.11 DCF  802.11 operates on DSSS, FHSS or IR PHY  MAC provides CSMA/CA through NAV (~’CS’)  Basic & RTS/CTS accesses  Congestion, timing and backoff mechanisms  On modeling DCF -> Bianchi; Wu, et. al.

4. A Problem with DCF Basic  2-way handshaking  Assumes that there is no other transmission during this slot!!!  What if there is a hidden station??? A BCD

5. Saturation Throughput Model  Bianchi provides a saturation throughput model based on a Markov model of backoff mechanism- P success E[P] P idle  + P success T s + P collision T c  P idle = 1- P tr and P success = P tr P s  P collision = P tr (1 - P s )  P tr = 1 – (1 –  ) n and P s = n  (1 –  ) n-1 /P tr  T s and T c measures time durations of a successful transmission and collided transmission S =

6. Hidden Station Model - Static  Kleinrock and Tobagi’s hearing graph- 1 1 1 0 0 1 21 1 0 0 1 3 0 0 1 1 1 4 0 0 1 1 0 5 0 1 0 0 1  Each station can hear some and not others => Pr(reachable) with assumption static => no transition  Generalize this to an n-station WLAN and decompose into a k-group reachability graph- P r(n) =  (N r(j) /N t(j) ) / k  Take average stations per group => expected number of hidden stations in the network 1, 23, 4 5 (a)(b)(c)

7. Hidden Station Model - Dynamic  Extend static model and allow transitions between k states, over n stations? => adjacency graph  Pr(reachable->reachable) => use control parameter,   Pr(hidden->*) = 1/l, Pr(reachable->hidden) = (1-  )/(l-1)  Balance equations: P r(j) + (1 – l) P h(j) = 1 (1 -  )/(1 - l) P r(j) = (1/l) P h(j)  Solve to get:P r(j) = 1 / (1 + l(1 –  )) 12 k 1 2 3 4 k-state Markov chain Adjacency graph

8. Our Throughput Model - Saturation  Worst case throughput loss => hidden stations always transmit  P tr = 1; P s = N re  (1 –  ) Nre-1  This changes throughput to- P s E[P]/(P s T s + P coll T c )  I also changed T c to include ACK_Timeout- DIFS+E[P]+SIFS+ACK_..  Huge degradation of throughput for either static or dynamic WLANs  Will see simulations agree

9. Our Throughput Model – Finite Load(1)  Similar grouping into k groups, but now with identical loads, i individually and   i = per group  Packet from a group must be successful both from its group and all other groups-  Further, transmission probabilities from k contending groups consisting some stations each  Plug P s and P tr into throughput equation  Can be used for both basic and RTS/CTS

10. Our Throughput Model – Finite Load(2)  Now have hidden groups, but assume same rate per group persists (i.e. allow only same rate within group)  Extend the previous P s and P tr to separate out reachable and hidden stations, in adjacency graph, i.e.,  Assumption that reachable >= hidden. Is it valid?  It is not obvious how to calculate . One idea may be from scheduler’s history at stations  Certainly justifies RTS/CTS, MACAW, DCF+, etc.

11. Simulation Topology & Traffic 1 2 3 4 5 <=250m >250m  Simulations in ns-2  914MHz Lucent WaveLAN DSSS PHY  Omni-antenna with 250m range  Modified CMU scene generator to create hidden stations, static topology, random pause time  Modified CMU traffic generator for variable packet size, intervals  RTS threshold => 3000 bytes  1028 bytes (8224 bits) packets  Inter-packet gap = 0 (saturation) and 1/rate (finite load)  CBR traffic over UDP links  Script to calculate various throughputs from trace

12. Saturation Simulation Results  Simulated with certain percentage hidden stations for 5, 10, 20, 50 stations  Results agree with model to some extent  Differences can be attributed to hidden stations may not always have packets (as assumed in the model)  Still need to experiment with  and simulate finite load throughput

13. Discussions & Conclusion  Hidden station models are sophisticated and can be used in many applications involving “carrier sense”  Saturation throughput model is valid and should be considered as an extension to Bianchi’s DCF model  Proposed finite load model is computationally expensive and needs further simplification. Finite load throughput model is an important step towards a general model of DCF and its derivatives  Though simulations are limited, it provides some degree of validation to the throughput models  It was a worthwhile investigation indeed helping me taking EE384* skills to different areas in networking

14. Summary & Future Work  Summarized prior art in DCF throughput and hidden station modeling  Developed static and dynamic hidden station models for 802.11 DCF  Developed a finite load throughput model for DCF  Integrated hidden station models for different types of loads  Showed limited simulation and …  Fixed relationships among reachable/hidden stations  Finite load validation with CBR traffic (per group)  Finite load validation with VBR traffic, e.g. Bernoulli IID, exponential, bursty,..  Scheduling packets in fixed src-dst pairs in multi- channel medium, e.g. iSLIP wireless networks

15. Q&A Simulation scripts, code, topologies, traffic pattern files can be found at- http://www.stanford.edu/~sirahman/80211dcf/ THANK YOU http://www.stanford.edu/~sirahman/80211dcf/