Skip to main content
Log in

Designing multihop wireless backhaul networks with delay guarantees

  • Published:
Wireless Networks Aims and scope Submit manuscript

Abstract

As wireless access technologies improve in data rates, the problem focus is shifting towards providing adequate backhaul from the wireless access points to the Internet. Existing wired backhaul technologies such as copper wires running at DSL, T1, or T3 speeds can be expensive to install or lease, and are becoming a performance bottleneck as wireless access speeds increase. Longhaul, non-line-of-sight wireless technologies such as WiMAX (802.16) hold the promise of enabling a high speed wireless backhaul as a cost-effective alternative. However, the biggest challenge in building a wireless backhaul is achieving guaranteed performance (throughput and delay) that is typically provided by a wired backhaul. This paper explores the problem of efficiently designing a multihop wireless backhaul to connect multiple wireless access points to a wired gateway. In particular, we provide a generalized link activation framework for scheduling packets over this wireless backhaul, such that any existing wireline scheduling policy can be implemented locally at each node of the wireless backhaul. We also present techniques for determining good interference-free routes within our scheduling framework, given the link rates and cross-link interference information. When a multihop wireline scheduler with worst case delay bounds (such as WFQ or Coordinated EDF) is implemented over the wireless backhaul, we show that our scheduling and routing framework guarantees approximately twice the delay of the corresponding wireline topology. Finally, we present simulation results to demonstrate the low delays achieved using our framework.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10
Fig. 11
Fig. 12
Fig. 13

Similar content being viewed by others

Notes

  1. 802.16 currently supports subchannelization in units of 1, 2, 4, 8, or 16 OFDM subcarriers in the uplink.

  2. Due to the spectral proximity of the subchannels the transmitted signal will drown out the received signal at an access point.

References

  1. Aguayo, D., Bicket, J., Biswas, S., Judd, G., & Morris, R. (2004, August). Link-level measurements from an 802.11b mesh network. In Proceedings of ACM SIGCOMM.

  2. Andrews, M., & Zhang, L. (1999, March). Minimizing end-to-end delay in high-speed networks with a simple coordinated schedule. In Proceedings of IEEE INFOCOM, pp. 380–388. New York, NY.

  3. Bao, L., & Garcia-Luna-Aceves, J. J. (2002, September). Transmission scheduling in ad hoc networks with directional antennas. In Proceedings ACM MobiCom, pp. 48–58, New York.

  4. Chambers, B. A. (2002, May). The grid roofnet: A rooftop ad hoc wireless network. Master’s thesis, Massachusetts Institute of Technology. http://www.pdos.lcs.mit.edu/grid/pubs.html.

  5. Chandra, R., Qiu, L., Jain, K., & Mahdian, M. (2004, October). Optimizing the placement of integration points in multi-hop wireless networks. In Proceedings of IEEE ICNP.

  6. Chlamtac, I., & Faragó, A. (1994). Making transmission schedules immune to topology changes in multi-hop packet radio networks. IEEE/ACM Transactions on Networking, 2(1), 23–29.

    Article  Google Scholar 

  7. Choudhury, R. R., Yang, X., Vaidya, N. H., & Ramanathan, R. (2002). Using directional antennas for medium access control in ad hocnetworks.In Proceedings of ACM MobiCom, pp. 59–70.

  8. Cole, R., Ost, K., & Schirra, S. (2001). Edge-coloring bipartite multigraphs in O(E logD) time. Combinatorica, 21, 5–12.

    Article  MATH  MathSciNet  Google Scholar 

  9. Dijkstra, E. W. (1959). A note on two problems in connection with graphs. Numerische Mathematik, 1, 269–271.

    Article  MATH  MathSciNet  Google Scholar 

  10. Erceg, V., Greenstein, L. J., et al. (1999). An empirically based path loss model for wireless channels in suburban environments. IEEE/ACM Journal on Selected Areas in Communications, 17(7), 1205–1211.

    Article  Google Scholar 

  11. Erceg, V., Hari, K. V. S., et al. (2001). Channel models for fixed wireless applications. IEEE 802.16 Broadband Wireless Access Working Group, IEEE 802.16.3c-01/29r4. http://ieee802.org/16.

  12. Gambiroza, V., Sadeghi, B., & Knightly, E. W. (2004). End-to-end performance and fairness in multihop wireless backhaul networks. In Proceedings of ACM MobiCom, pp. 287–301.

  13. Garey, M. R., & Johnson, D. S. (1979). Computers and intractability: A guide to the theory of NP-completeness. San Francisco, CA: W.H. Freeman and Co.

    MATH  Google Scholar 

  14. Huang, X. L., & Bensaou, B. (2001). On max-min fairness and scheduling in wireless ad-hoc networks: Analytical framework and implementation. In Proceedings of ACM MobiHoc, pp. 221–231.

  15. IEEE Computer Society and IEEE Microwave Theory and Techniques Society. (2004, October). 802.16-2004: Air Interface for Fixed Broadband Wireless Aceess Systems. IEEE Standards.

  16. Jain K., Padhye, J., Padmanabhan, V. N., & Qiu, L. (2003). Impact of interference on multi-hop wireless network performance. In Proceedings of ACM MobiCom, pp. 66–80.

  17. Ju, J.-H., & Li, V. O. K. (1998). An optimal topology-transparent scheduling method in multihop packet radio networks. IEEE/ACM Transactions on Networking, 6(3), 298–306.

    Article  Google Scholar 

  18. Kanodia, V., Li, C., Sabharwal, A., Sadeghi, B., & Knightly, E. (2001). Distributed multi-hop scheduling and medium access with delay and throughput constraints. In Proceedings of ACM MobiCom.

  19. Karrer, R., Sabharwal, A., & Knightly, E. (2003, November). Enabling large-scale wireless broadband: The case for TAPs. In 2nd Workshop on Hot Topics in Networks (HotNets-II), Cambridge, MA.

  20. Kodialam, M., & Nandagopal, T. (2003, August). Characterizing the achievable rates in multihop wireless networks. In Proceedings of ACM MobiCom, San Diego, CA.

  21. Koffman, I., & Roman, V. (2002). Broadband wireless access solutions based on OFDM access in IEEE 802.16. IEEE Communications Magazine, 40(4), 96–103.

    Article  Google Scholar 

  22. König, D. (1916). Graphok és alkalmazásuk a determinánsok és a halmazok elméletére (in Hungarian). Mathematikai és Természettudományi Értesítő, 34, 104–119.

    Google Scholar 

  23. Le Boudec, J. Y., & Thiran, P. (2004). Network calculus. Springer Verlag. http://ica1www.epfl.ch/PS_files/NetCal.htm.

  24. Li, C., & Knightly, E. (2000, November). Coordinated network scheduling: A framework for end-to-end services. In Proceedings of IEEE ICNP, Osaka, Japan.

  25. Luo, H., Cheng, J., & Lu, S. (2004). Self-coordinating localized fair queueing in wireless ad hoc networks. IEEE/ACM Transactions on Mobile Computing, 3(1), 86–98.

    Article  Google Scholar 

  26. Luo, H., Lu, S., & Bharghavan, V. (2000). A new model for packet scheduling in multihop wireless networks. In Proceedings of ACM MobiCom, pp. 76–86.

  27. Mirchandani, R., & Francis, R. (Eds.). (1990). Discrete location theory. New York, NY: Wiley.

    MATH  Google Scholar 

  28. Parekh, A. K., & Gallager, R. G. (1994). A generalized processor sharing approach to flow control in integrated services networks: The multiple-node case. IEEE/ACM Transactions on Networking, 2(2), 137–150.

    Article  Google Scholar 

  29. Ramanathan, S. (1997). A unified framework and algorithm for channel assignment in wireless networks. In Proceedings IEEE INFOCOM, Kobe, Japan.

  30. Ramanathan, R. (2001, October). On the performance of ad hoc networks using beamforming antennas. In Proceedings of ACM MobiHoc, Long Beach, CA.

  31. Viswanathan, H., & Mukherjee, S. (2003). Performance of cellular networks with relays and centralized scheduling. In Vehicular Technology Conference, VTC 2003-Fall, pp. 1923–1928.

  32. WiMAX Forum. (2004, October). Business case models for fixed broadband wireless access based on WiMAX technology and the 802.16 standard. Whitepaper. http://www.wimaxforum.org.

Download references

Acknowledgement

The authors wish to thank Matthew Andrews for many helpful discussions.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Lisa Zhang.

Appendices

Appendix A: hardness proof for routing

A feasible routing was defined in Sect. 4.

Theorem 4.1

Given the link capacities and the connection bit rates, determining if there is a feasible routing is NP-complete.

Proof

Clearly checking if a routing is feasible can be done in polynomial time. We show that the problem is NP-hard by a reduction from the known NP-complete problem called 3-Partition [13]. An instance of 3-Partition consists of a set A of 3m elements, a positive integer B and a positive integer size s(a) for each ∈ A where B/4 < s(a) < B/2 and such that ∑aA s(a) = mB. The goal is to find a partition of A into m disjoint sets \(A_1, A_2,\ldots,A_m\) such that for 1 ≤ ≤ m, \(\sum\nolimits_{a\in A_{i}}s(a) =B.\) Notice that if such a partition exists, then each A i contains exactly 3 elements of A. This problem is NP-complete in the strong sense.

Consider an instance I of 3-Partition. We now describe an instance J of the feasible routing problem based on I that can be constructed in time polynomial in I and such that I has a solution if and only if J does.

The construction is basically shown in Fig. 14. There is a root node R, a node labeled a i for each a i  ∈ A, nodes labeled \(A_1, A_2,\ldots,A_m\) representing a partition of A and dummy nodes labeled \(D_1,D_2,\ldots, D_m.\) Each node a i has a connection in the up link direction with rate s(a i ). These are the only connections.

Fig. 14
figure 14

Illustrating the NP-completeness construction

Links of the form (a i , A j ) or (A j D j ) have capacity 2B while links of the form (D j , R) have capacity 2mB. All other links (not shown) have capacity 0.

Suppose more connections with total rate greater than B are routed through A j . Then the congestion at A j due to the link (A j , D j ) is more than B/2B = 1/2 and hence this would not be a feasible routing. Thus for any feasible routing each A j must have connections routed through it with total rate exactly B. It’s easy to check that any routing that results in every node A j have connections with total rate B routed through it is a feasible routing. Hence this feasible routing instance has a solution if and only if the 3-Partition instance has a solution.\(\hfill\square\)

Appendix B: propagation delay

We use the framework of GR server in [23] to prove Lemma 5.5 for WFQ. Let us first define guaranteed rate (GR) server.

Definition 2 (Definition 2.1.1 [23])

Consider a server that serves a connection. Packets are numbered in the order of arrival. Let a n  ≥ 0 and d n  ≥ 0 be the arrival and departure times of the n packet. A server is called a guaranteed rate server for this flow with rate r and latency ℓ if it guarantees that d n  ≤ f n  + ℓ where f n is defined below and L is the packet size of the connection.

$$ \left\{\begin{array}{llll} f_0&=&0&\\ f_n&=&\max\{a_n,f_{n-1}\}+L/r & \forall n > 0 \end{array}\right. $$

Fact B.1

Consider a server with capacity c and a set of connections C that go through the server. Let\(R_i=c\cdot \frac{\rho_i}{\sum\nolimits_{j\in C} \rho_j}\)for connection i ∈ C,

  1. 1.

    Under the GPS scheduler the server is a GR server with rateR i and latency 0 for connectioni.

  2. 2.

    Under the WFQ scheduler the server is a GR server with rateR i and latencyLmax/c for connectioni, whereLmax is the maximum packet size among all connections inC.

  3. 3.

    Under the WFQ scheduler the server with propagation delayg is a GR server with rateR i and latencyg + Lmax/c for connectioni.

We note that Fact B.1.1 follows from the definition of the GPS protocol. Fact B.1.2 follows from Proposition 2.1.1 in [23]. Fact B.1.3 says that propagation delay defers the finishing time by g.

Theorem B.1 (Theorems 1.4.2, 1.7.2 and 2.1.3 [23])

If a GR server has rate r and latency for connection i then the delay bound for this connection is

$$ \frac{\sigma_i}{r} + \ell, $$

where σ i is the burst size of connection i.

Theorem B.2 (Theorem 2.1.5 [23])

ConsiderK i servers where each server 1 ≤ ≤ K i is GR with ratesr k and latencies k for connectioni. The concatenation of theseK i servers is GR with rater = min k r k and latency\(\ell=\sum\nolimits_{1\le k\le K_{i}} \ell_k +\sum\nolimits_{1\le k\le K_{i}-1} L_i/r_k\)for connectioni.

The combination of Theorem B.1 and Theorem B.2 allows us to derive the following delay bound for K i concatenated GR servers.

$$ \frac{\sigma_i}{\min_{1\le k\le K_{i}} r_k}+\sum_{1\le k\le K_{i}} \ell_k + \sum_{1\le k\le K_{i}-1} \frac{L_i}{r_k} $$
(9)

Under WFQ, for connection i each server k has rate \(r_k=c_k\cdot \frac{\rho_i}{\sum_{j} \rho_j}\) and latency \(\ell_k=L_{\max}/c_k\) by Fact B.1.2 where c k is the capacity of serve k. Due to the admissibility of the traffic, we have r k  ≥ ρ i for all i and k which implies ≥ ρ i . Therefore, Eq. 9 implies the end-to-end delay under WFQ with zero propagation is

$$ \frac{\sigma_i}{\rho_i}+\sum_{1\le k\le K_{i}} \frac{L_{\max}}{c_k} + (K_i-1)\cdot \frac{L_i}{\rho_i}. $$

Note that the above bound is given in [28]. By Fact B.1.3, we have \(\sum\nolimits_{1\le k\le K_{i}}\ell_k = g(i)+\sum\nolimits_{1\le k\le K_{i}}L_{\max}/c_k\) where g(i) is the total propagation delay along the path for connection i. Therefore, the end-to-end delay under WFQ with propagation is

$$ \left(\frac{\sigma_i}{\rho_i}+\sum_{1\le k\le K_{i}} \frac{L_{\max}}{c_k} + (K_i-1)\cdot \frac{L_i}{\rho_i}\right)+g(i). $$

Therefore, Lemma 5.5 holds for WFQ.

The Coordinated Earliest Deadline First (CEDF) scheduling protocol was described and analyzed in [2] for zero propagation delay. Consider a packet p from connection i. In [2] the authors define the following deadline D k , 1 ≤ ≤ K i , for each link that p goes through. When multiple packets compete for a link, the scheduler gives priority to the packet that has the earliest deadline. The schedule is “coordinated” since the deadlines are incremented along the way.

$$ \left\{\begin{array}{ll} D_1 = \tau+ G_{{m_{1} }} \\ D_k = D_{k-1}+ G_{{m_{k} }} \end{array}\right. $$
(10)

Roughly speaking, τ is a random number which serves the purpose of spreading out packets that are injected in a burst. The value of τ is upper bounded by \(t_{inj}+O((\sigma_i+L_i)/\rho_i)\) where t inj is the injection time of packet p. The value of \(G_{{m_{k} }} \) is proportional to \(L_{\max}/c_{{m_{k} }}\) where m k is the kth link along the path for connection i. (See [2] for the detailed definition of τ and \(G_{{m_{k} }}. \)) To prove a delay bound for CEDF, [2] shows the following two lemmas.

Lemma B.3 (Lemma 2 [2])

With high probability, for everyt that is a potential deadline the time interval [\(t - G_{{m_{k} }} ,t\)] has at most\(G_{{m_{k} }} \)bits of packets that have deadlines within the interval at linkm k .

Lemma B.4 (Lemma 3 [2])

If the condition in Lemma B.2 holds, then every packet meets all its deadlines.

These two lemmas immediately imply that the end-to-end delay is \(\tau-t_{inj}+\sum\nolimits_k G_{{m_{k} }}.\) Combined with the definitions of τ and \(G_{{m_{k} }}, \) [2] states

Theorem B.5 (Theorem 1 [2])

With high probability, the end-to-end delay under CEDF is upper bounded by

$$ \frac{\sigma_i+4L_i/\varepsilon}{\rho_i} + \sum_{1\le k\le K_{i}}\frac{L_{\max}}{ c^{{m_{k} }}}\log(\cdot). $$

If link e has propagation delay g e , we redefine the deadlines as follows.

$$\begin{aligned} \left\{ \begin{array}{ll} D_1 = \tau+G_{{m_{1} }}+g_{{m_{1} }}\\ D_k = D_{k-1}+G_{{m_{k} }}+g_{{m_{k} }} \end{array} \right.\end{aligned}$$
(11)

That is, each deadline is postponed by the corresponding propagation delay. Both Lemmas B.3 and B.4 hold for identical proofs under the new definitions (11) of the deadlines. Therefore, with link propagation delays the end-to-end delay becomes \(\tau-t_{inj}+\sum\nolimits_k G_{{m_{k} }}+\sum\nolimits_k g_{{m_{k} }}.\) In other words, if connection i has total propagation delay of g(i) its delay bound under CEDF becomes,

$$ \left(\frac{\sigma_i+4L_i/\varepsilon}{\rho_i} + \sum_{1\le k\le K_{i}}\frac{L_{\max}}{ c^{{m_{k} }}}\log(\cdot)\right) + g(i).$$
(12)

Lemma 5.5 holds for CEDF.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Narlikar, G., Wilfong, G. & Zhang, L. Designing multihop wireless backhaul networks with delay guarantees. Wireless Netw 16, 237–254 (2010). https://doi.org/10.1007/s11276-008-0127-0

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11276-008-0127-0

Keywords

Navigation