The Illusion of Being Deterministic – Application-Level Considerations on Delay in 3G HSPA Networks

Abstract

Abstract. The delay experienced by mobile applications in HSPA networks depends to a large extent on highly dynamical global context like, e.g., cell load or algorithms and thresholds governing radio resource scheduling, and on local context like, e.g., user-generated load or load history. These complex uncertainty factors are outside of an applications sphere of influence and result in applications perceiving HSPA link behavior as non-deterministic and non-reproducible.This paper analyzes accurate round-trip and one-way delay measurement results for three public HSPA networks to demonstrate the high degree of network non-determinism which mobile applications are likely to encounter in practice, particularly significant payload-dependence and halved delay on slightly increased user-generated link load. We argue that current HSPA radio link schedulers, relying on instantaneous user load as decision criterion for channel capacity allocation, neglect real-time application requirements. Cross-layer optimization is one solution which enables deterministic scheduler decisions based on application requirements.

Representative Delay Measurements (RDM): Facing the Challenge of Modern Networks

“…Main factor which decides on whether a measurement packet is subject to lower or higher delay is preliminary state, in particular data rate and interpacket interval. Earlier work [9] shows that mobile operators use different scheduling-and allocation policy, such that measurement results and diagrams for distinct mobile networks differ substantially for the same scenario, i.e., identical traffic. Fig.…”

“…Various authors have published results of active and passive delay measurements in mobile cellular networks, including [6], [7], and [8]. Earlier work ( [9], [10]) has presented methodological drawbacks of measurements in mobile cellular networks and compared HSPA delay figures for several HSPA network vendors and operators. Recent publications propose delay measurement techniques in reactive networks [11] and present a detailed discussion on theoretical and practical consequences of the randomness cancellation effect in time-slotted networks [12].…”

Mobile SIP: An Empirical Study on SIP Retransmission Timers in HSPA 3G Networks

“…The contribution of different network components (user equipment, NodeB and RNC) on HSUPA end-to-end delay is analyzed in [10], inferring on 30 ms delay for 1 kbyte packets. In earlier work ( [12] and [13]) we have dissected difficulties and uncertainty factors to be expected when measuring mobile network delay, focusing on randomness in terms of inter-packet delay and payload size as key factor to reveal load-dependent network behavior.…”

Dissecting 3G Uplink Delay by Measuring in an Operational HSPA Network

“…Otherwise the network scheduler would release the HSUPA connection and force the modem to WCDMA Forward Access Channel (FACH) operation, in order to save radio resources. Consequences of such a fallback are observed in [4] and [3], resulting in very high delay values for small packet sizes. In the context of this study these effects are undesired and hence avoided.…”

“…Their traffic generation method differs significantly from ours. The authors of [4] and [5] provide OWD measurements with low timestamping accuracy from multiple network operators. They use ICMP ping messages as measured data traffic, in order to highlight the importance of the right data generation method, which has to be RCF 2330 [6] compliant in their opinion.…”