Ivan N. Vukovic

Performance analysis of the random access channel (RACH) in WCDMA

In addition to call origination and registration messages, the random access channel (RACH) in 3GPP will carry traffic including short messaging service (SMS) packets and possibly short data bursts in the absence of a dedicated traffic channel. Because of this increase in traffic volume it is important to accurately predict performance of this channel and size the resources appropriately. It has commonly been assumed that RACH throughput is comparable to that of slotted ALOHA. Using the infinite user assumption we show that the resource utilization at node B is significantly higher than e/sup -1/. This analysis provides a close estimate of the RACH capacity when the actual protocol is in place. We present physical layer results that provide preamble detection probability at the receiver and use these results in a detailed MAC layer simulation with finite number of mobiles. We analyze the capacity per resource in node B for two message sizes and we investigate performance impacts of power capture and the maximum number of preamble cycles.

Delay analysis of different backoff algorithms in IEEE 802.11

We use a newly developed analytical model to compute the average delay of the existing binary exponential backoff (BEB) algorithm in IEEE 802.11 wireless LAN and two different proposals, exponential increase exponential decrease (EIED) and exponential increase linear decrease (EILD) backoff algorithms. A one-dimensional Markov chain model is constructed for each algorithm and used to compare delay under saturation conditions. Our approach simplifies previous analyses of the BEB algorithm which used a 2-dimensional Markov chain model, while the other backoff algorithm delays have not been analyzed before. Additionally, improvements have been made to the previous analysis of BEB to achieve more accurate results. Analytical results are compared to those obtained from simulation.

Saturation throughput analysis of different backoff algorithms in IEEE802.11

We develop a new analytical model of the existing binary exponential backoff (BEB) algorithm in IEEE 802.11 wireless LAN and two different proposals, exponential increase exponential decrease (EIED) and exponential increase linear decrease (EILD) backoff algorithms. A one-dimensional Markov chain model is constructed for each algorithm and used to compare throughput under overload conditions. Our approach simplifies previous analyses of the BEB algorithm which used a 2-dimensional Markov chain model. The saturation throughput as well as average backoff window and probability of success for all three algorithms are calculated and compared in various types of traffic and network loads. Analytical results are compared to the simulation results. We found that EIED performs always better than BEB, while EILD provides improvement at higher loads. Additionally, we show that the infinite retransmission assumption provides a very tight upper bound on throughput for all schemes.

Throughput comparison of random access schemes in 3GPP

Random access channel (RACH) and common packet channel (CPCH) are the two common link transport channels in 3GPP based on random access. Although their design goals were different both access schemes could be used for short message transport. To assist the system engineering of data services in UMTS it is of interest to compare performance of the two schemes. It has commonly been assumed that the normalized RACH throughput is comparable to that of slotted ALOHA and thus bounded by 36%. Additionally, it has been assumed that the collision probability is RACH is significantly higher than in CPCH due to collision detection phase introduced in the latter. In this paper we challenge these two commonly used assumptions. Using the same method we derive the probability of successful transmission for both RACH and CPCH. We then show analytically for a single demodulation resource at node B and under certain set of assumptions, that the maximum normalized throughput is greater than 80% in both cases. We make two main observations: the throughput advantage for CPCH is mainly due to the longer message size and the effect on throughput of the CD phase is not significant. For multiple resources we use simulations to obtain normalized throughput per channel resource.

Spectrum sharing under the asynchronous UPCS etiquette: the performance of collocated systems under heavy load

Recently the FCC opened three 10 MHz bands for unlicensed use. In order to operate in UPCS bands, devices must comply with rules known as the UPCS etiquette (United States Code of Federal Regulations, Title 47, Part 15(d)). In this paper we study channel sharing between two or more collocated systems under the asynchronous UPCS etiquette. In particular we show that under heavy load individual systems have a tendency to hold the channel for hundreds of milliseconds, thus blocking all traffic in other, competing systems. We have calculated the distribution of the blocking time for two versions (or interpretations) of the UPCS etiquette. The impact of the average blocking time on delay sensitive traffic is discussed and possible improvements achieved through a tradeoff between system capacity and average blocking time are investigated.

An RTS-on-demand Mechanism to Overcome Self-interference in an 802.11 System

The IEEE 802.11 standard defines a Request to Send/Clear to Send (RTS/CTS) mechanism to reduce the impact from self-interference such as hidden terminals. However, the usage of RTS/CTS is only based on a pre-defined fixed packet size threshold, focusing only on the cost rather than the presence of interference. This paper presents an RTS on Demand (RoD) mechanism to dynamically utilize the RTS/CTS based on the interference situation (e.g. collision) by monitoring various parameters such as response frame to an unknown station, retransmission attempt, and collision rate monitored at the intended receiver. Numerical analysis of the proposed RoD mechanism compared to the conventional 802.11 RTS/CTS mechanism is provided in the paper.

PPPmux-a new protocol for transporting small IP packets

We introduce a new protocol to improve the capacity of communication links to carry real-time applications that generate a large number of small packets. The key issue in the transport of such applications is the relatively large protocol overhead for each packet. For example, the overhead for RTP/UDP/IP/PPP/HDLC is 12/8/20/7 bytes respectively resulting in a total of 47 bytes. This is considerable for applications with a small payload (10-20 bytes). RTP/UDP/IP header compression (cRTP) mitigates the problem partially by reducing the corresponding protocol overhead from 40 bytes to a best case of 2 bytes. PPPmux provides a further improvement by multiplexing multiple application-level packets into a single PPP frame thereby reducing the PPP overhead per packet. One potential deployment of PPPmux is in cellular networks (IS-95A and UMTS) on links connecting base stations to the radio access network controllers (RNC). The paper provides numerical results listing capacity improvements over existing schemes. Since, multiplexing concatenates multiple IP packets into a single PPP frame, a frame loss would result in loss of multiple IP packets. We provide guidelines on the choosing the optimal multiplexing size that gives the best trade-off between increase in link capacity and increase in packet loss. Specifically, the multiplexing size of 200 bytes gives the best compromise between capacity and packet loss for typical cellular applications.

HIPERLAN type I: performance analysis of the channel access control protocol

The high performance radio LAN (HIPERLAN) type 1 specification was adopted as a standard by the European Telecommunications Institute (ETSI) in late 1996. While many contributions on the performance of wireless LANs had been submitted in the standardization effort, certain HIPERLAN performance issues were left unresolved. Most of the published performance work so far has been based on simulations, not the analytic results. In this paper we present an analysis of HIPERLAN channel access control protocol in an ad hoc networking scenario. We derive expressions for the probability of success, the average overhead per-channel access cycle, and the network capacity. Using an M/G/1 queueing model we approximate the average delay using equilibrium point analysis techniques. In addition, a lower and an upper bound on the average delay as a function of the throughput are obtained.

CDMA 1X radio access network IP backhaul sizing analysis

This paper focuses on the sizing methodology for the link between the base station transceiver (BTS) and the central base station controller (CBSC). Packet-switched technology has promised greater flexibility and increased utilization of resources for the cellular operators. This brought new challenges in terms of traffic engineering that could not be addressed by Erlang-B formula alone. We present analysis techniques for IP backhaul sizing in CDMA 1X radio access network (RAN). We first present the 2G/2.5G circuit backhaul efficiency analysis for voice traffic. Using the central limit theorem we analyze 3G packet-switched backhaul efficiency with mixed voice, data, and signaling traffic for a single T1 link. Applying the knowledge gained from the single backhaul link efficiency analysis, we present practical steps to calculate the number of backhaul links required to meet QoS demands given a operator-specified call model.

Impact of capture and varying transmit power levels on saturation throughput in IEEE 802.11

Multiple backlogged stations communicating with a central receiving node or access point using IEEE 802.11 MAC is analyzed. When packets are sent simultaneously often one of the packets can be successfully received by the access point due to power capture effect. Several schemes for assigning transmit power level to each station are investigated along with two capture models. We apply a well know saturation throughput methodology developed by Bianchi (2000) which was recently simplified allowing schemes other than binary exponential backoff to be analyzed. With stations in different radio conditions the user population is not homogeneous and the collision probability is not fixed. Two schemes are investigated in which each station is either assigned a fixed power level from a set of & levels or randomly chooses a power level for each transmission from the same set. We show that power capture improves performance of 802.11 in the basic mode for each scheme. We also calculate individual user throughput and show that significant fairness imbalance arises with fixed power assignments. We also provide an upper bound for the throughput assuming perfect knowledge of the number of colliding stations