Академический Документы
Профессиональный Документы
Культура Документы
Abstract— Uplink scheduling in wireless systems is gaining im- In this work, we consider the problem of optimal scheduling
portance due to arising uplink intensive data services (ftp, image of uplink user transmissions in a single CDMA cell. We assume
uploads etc.), which could be hampered by the currently in-built that the system operates in a TD/CDMA manner, with time-
asymmetry in favor of the downlink. In this work, we propose and
study algorithms for efficient uplink packet-data scheduling in a slotted scheduling of transmissions, assisted by periodic feed-
CDMA cell. The algorithms attempt to maximize system through- back of channel and/or congestion information through control
put under transmit power limitations on the mobiles assuming in- channels. The base station then controls user transmission rates
stantaneous knowledge of user queues and channels. However no via downlink signaling. While our general goals are similar to
channel statistics or traffic characterization is necessary. Apart those of the prior work on uplink scheduling, we formulate the
from increasing throughput, the algorithms also improve fairness
of service among users, hence reducing chances of buffer overflows problem in a different manner that enables efficient computa-
for poorly located users. tion of throughput optimal schedules. Moreover, we observe
The major observation arising from our analysis is that it is that the optimal schedules are reasonably well approximated by
advantageous on the uplink to schedule “strong” users one-at-a- some very simple scheduling rules that may be implemented
time, and “weak” users in larger groups. This contrasts with with modest effort in 3G systems. The optimal schedules, as
the downlink where one-at-a-time transmission for all users has
shown to be the preferred mode in much previous work. Based on well as our proposed approximations, seem to possess an in-
the optimal schedules, we propose less complex and more practi- tuitively appealing property that is consistent with traditional
cal approximate methods, both of which offer significant perfor- observations about wireless transmission: it is advantageous
mance improvement compared to one-at-a-time transmission, and for users with weak channels to transmit simultaneously, and
the widely acclaimed Proportional Fair (PF) algorithm, in simu- for users with strong channels to transmit one-at-a-time1 . The
lations. When queue content cannot be fed back, we propose a
simple modification of PF, Uplink PF (UPF), that offers similar intuition behind this is that the added interference at the base
improvement. from simultaneous transmissions of weakly recieved users to
each other is small compared to the extraneous interference,
Index Terms—Scheduling, Uplink, Reverse Link, CDMA.
and thereby does not affect the user SIRs and data rates signifi-
cantly. On the contrary, for users recieved strongly at the base,
the penalty in terms of SIR and data rate with simultaneous
I. I NTRODUCTION
transmission can be quite significant.
Data scheduling in wireless networks is a widely studied As mentioned earlier, there is some previous work on the
topic due to the impending explosion of high speed wire- subject of this paper, scheduling for CDMA uplink [7], [10],
less data services in third generation (3G) systems. For nat- [9]. However, the previous work addresses somewhat different
ural reasons associated with the expected traffic characteris- problems, and does not offer the solutions that we do below to
tics, most of the previous research has focused on the forward- maximize uplink scheduling gains while meeting rate/QoS re-
link/downlink, i.e. base to mobile communication (see [1], quirements among competing users. Perhaps the closest work
[2], [3] and references therein for examples, or proceedings of that arrives at roughly similar qualitative conclusion of “One-
the last several IEEE INFOCOMs!). The traffic is expected to strong or many-weak” is [13], which solves a dynamic pro-
be dominated by web browsing and file downloads. As a re- gramming formulation of a static uplink scheduling problem.
sult, current wireless data systems employ highly asymmetric Some useful observations regarding the optimality of “bang-
link designs (e.g. HDR) with skinny uplinks and fat downlink bang” control (each user is either silent or transmiting at full
pipes [4]. However, it has also often been pointed out that there power in each slot) were made in [14] and subsequent work,
could be a proportional increase in reverse-link/uplink traffic but the authors do not solve the single-cell optimal scheduling
in the form of acknowledgments, feedback etc. along with problem conclusively. The approach of maximizing rate sum
the growth of other services like ftp, image/data uploads etc. or SIR sum, taken in [10], [9] is often a practically poor option,
which require high data rates on the uplink. These considera- since it mostly benefits users in favorable locations while ignor-
tions have resulted in some research on the subject of uplink ing the performance seen by the disadvantaged users. As for
scheduling [5], [6], [7], [8], [9], [10], although small com- the information theoretic approaches of [11], [12], these works
pared to the literature on downlink scheduling. There is also deal extensively with various issues such as optimal power con-
some amount of related work from information theoretic point trol, ergodic capacity, delay limited capacity etc., but in all
of view, see [11], [12] and references therein, on the subject of cases with constraints on long term average transmit power and
wireless multiaccess channels. We discuss the relationship of assuming continuously backlogged users with known channel
our work to these approaches in greater detail below.
1 Hereafter, we refer to users with low recieved power at the base even when
The authors are from the Math Research at Bell Labs, Lucent Tech., Murray transmitting at peak transmit power as “weak” users, and the strongly recieved
Hill, NJ. E-mail: {kumaran,ljqian}@research.bell-labs.com. users at the base as “strong” users.
subject to
N
αi
αj + ≤1 , αi ≥ 0 ∀i. (9) Fig. 1. Rate vs. α for f ∈ [0.5, 1] for fixed Λ.
j=1
P̄i
For any fixed value of Λ, (10) and (11) have a simple greedy
Typically gi (·) = g(·) ∀ i are identical functions, as is the case solution for gi (·) convex. The idea behind the procedure is il-
for the Shannon formula, but our results remain unaffected even lustrated in Figure 1, where the convex rate curve is replaced
if they were all different, as long as they stay convex. Before by a straight line joining the endpoints. This is valid on account
discussing methods to solve (8), (9), we observe some useful of Theorem 1, which implies that the αi always take one of the
properties of the optimal solution. bounding values in the constraints of eqs. (11) if Λ is appropri-
Theorem 1: The optimal schedule has the property that each ately chosen. The algorithm to construct the solution is outlined
transmitting user transmits at full power, i.e. Pi = 0 for some below:
subset S of the users and Pi = P̄i for the complementary set S̄. Algorithm OPT:
1) Let ᾱi = min{Λ, P̄i (1 − Λ)}. Order the users according
Proof: Equations (9) specify 2N constraints on the feasi- to decreasing value of the quantity vi = w ᾱi gi (ᾱi ).
i
ble αi . From standard theorems on convex maximization with 2) Assign αi = ᾱi for the user with the highest value of vi .
linear constraints, it is easy to see that the optimum occurs at 3) Update as Λ → Λ − ᾱi and repeat with remaining users.
corner point of (9) due to the joint-convexity of (8) in the αi . The user ordering will not change if Λ ≥ P̄i (1 − Λ) for
Pi
≥ γi . (12) The optimal scheduling methods discussed earlier continue
1 + f j∈{1,...,N },j=i Pj
to apply for f ∈ [0.5, 1] since g(·) is still convex in this range
It is easy to see that most of the previous analysis goes through for the Shannon rate formula. In the following, we address the
unchanged if one sets γi → f γi and (Pi , P̄i ) → (f Pi , f P̄i ). problem of optimal scheduling, as stated in (14), (15), for f ∈
10
0.3 8
0.25 6
0.2
4
0.15
2
0.1
0.5 0.55 0.6 0.65 0.7 0.75 0.8
0
(a) Percentage of ‘‘weak’’ Users 20 25 30 35 40 45 50
(a) Traffic Arrival Rates (kbps per user)
0.25 5000
MaxQR
4500 PF
UPF
QRP
0.2 4000 OPT
0.15 3000
2500
0.1 2000
1500
0.05 1000
500
0 0
0 2 4 6 8 10 12 14 16 18 20 20 25 30 35 40 45 50
(b) Number of simultaneously transmitting users (b) Traffic Arrival Rates (kbps per user)
4
x 10
Distribution of Average Rate per user per slot
9
0.1
’t’
MaxQR
8 PF
0.09 UPF
QRP
7 OPT
0.08
0.07
6
Queue Length (bits)
0.06 5
Fraction of Total Slots
0.05 4
0.04
3
0.03
2
0.02
1
0.01
0
20 25 30 35 40 45 50
0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 (c) Traffic Arrival Rates (kbps per user)
(c) Avg. Rate (Kbps)
Fig. 5. (a) Total average queue length (b) Queue length of a typical “strong”
Fig. 4. (a) Performance gain of QRP over MaxQR vs. percentage of “weak” user (c) Queue length of a typical “weak” user, applying MaxQR, PF, UPF,
mobile users (b) Distribution of simultaneously transmitting users. (c) Distri- QRP and OPT vs. offered traffic.
bution of avg. rate per user over slots.
VIII. C ONCLUSION
We have proposed optimal uplink scheduling algorithms for a
single CDMA cell, and related efficient approximate algorithms
for practical implementation. Simulations demonstrate substan-
tial performance improvement with these algorithms, and we
also discuss the implementation requirements they entail. Fur-
ther research is required to address multi-cell systems, and also
to incorporate the effects of soft-handoff, which may have a
significant influence in uplink scheduling.
R EFERENCES
[1] N. Kahale and P. E. Wright, “Dynamic Global Packet Routing in Wireless
Networks,” in Proc. of INFOCOM’97, 1997.
[2] P. Bender et al, “CDMA/HDR A Bandwidth Efficient High Speed Wire-
less Data Service for Nomadic Users,” IEEE Communications Magazine,
2000.
[3] D. M. Andrews, K. Kumaran, K. Ramanan, A. Stolyar, R. Vijayakumar,
and P. Whiting, “CDMA Data QoS Scheduling on the Forward Link with
Variable Channel Conditions,” Bell Labs Technical Memo No.10009626-
000404-05TM, 2000.
[4] E. Esteves, “On the Reverse Link Capacity of CDMA2000 High Rate
Packet Data Systems,” in Proc. of ICC’02, may 2002, vol. 3.
[5] S.V. Hanly, Information capacity of Radio Networks, Ph. D. Thesis,
Cambridge University, England, 1993.
[6] A. Sampath and J. Holtzman, “Access Control of Data in Integrated
Voice/Data CDMA Systems: Benefits and Tradeoffs,” IEEE Journal on
Selected Areas in Comm., vol. 15, no. 8, oct. 1997.
[7] S. Ramakrishna, Optimal Scheduling of CDMA Systems, Ph. D. Thesis,
WINLAB, Rutgers University, Piscataway, NJ, 1996.
[8] L. Qian, Optimal Power Control in Cellular Wireless Systems, Ph. D.
Thesis, Rutgers University, Piscataway, NJ, 2001.
[9] C. W. Sung and W. S. Wong, “Power Control and Rate Management for
Wireless Multimedia CDMA Systems,” IEEE Trans. on Comm., vol. 49,
no. 7, july 2001.
[10] S. Ulukus and L. J. Greenstein, “Throughput Maximization in CDMA
Uplinks Using Adaptive Spreading and Power Control,” in Proc. of IEEE
6th Intl. Symposium on Spread-Spectrum Tech. and Appl., 2000.
[11] S. Hanly and D. Tse, “Multiaccess Fading Channels - Part II: Delay
Limited Capacities,” IEEE Trans. on Info. Theory, vol. 44, no. 7, nov.
1998.