[Paper Review] Capacity and Scheduling of Access Points for Multiple Live Video Streams

I-Hong Hou, Rahul Singh|arXiv (Cornell University)|Jun 10, 2013

Advanced Wireless Network Optimization7 references8 citations

TL;DR

This paper proposes the Earliest Positive-Debt Deadline First (EPDF) scheduling policy to maximize the throughput and minimize per-packet delay for multiple live video streams over unreliable wireless links. By leveraging buffering at the server and dynamically prioritizing packets based on residual delay and throughput debt, EPDF achieves the theoretical capacity region, outperforming existing policies by a large margin in trace-based simulations across three video traces.

ABSTRACT

This paper studies the problem of serving multiple live video streams to several different clients from a single access point over unreliable wireless links, which is expected to be major a consumer of future wireless capacity. This problem involves two characteristics. On the streaming side, different video streams may generate variable-bit-rate traffic with different traffic patterns. On the network side, the wireless transmissions are unreliable, and the link qualities differ from client to client. In order to alleviate the above stochastic aspects of both video streams and link unreliability, each client typically buffers incoming packets before playing the video. The quality of the video playback subscribed to by each flow depends, among other factors, on both the delay of packets as well as their throughput. In this paper we characterize precisely the capacity of the wireless video server in terms of what combination of joint per-packet-delays and throughputs can be supported for the set of flows, as a function of the buffering delay introduced at the server. We also address how to schedule packets at the access point to satisfy the joint per-packet-delay-throughput performance measure. We test the designed policy on the traces of three movies. From our tests, it appears to outperform other policies by a large margin.

Motivation & Objective

• Address the challenge of delivering multiple live video streams over unreliable wireless links with variable-bit-rate traffic and heterogeneous channel conditions.
• Characterize the precise capacity region in terms of joint per-packet delay and throughput constraints under controlled buffering at the server.
• Design an online scheduling policy that guarantees performance within the capacity region without requiring knowledge of future traffic patterns.
• Evaluate the tradeoffs between long-term fairness and short-term performance under realistic video traffic and channel dynamics.

Proposed method

• Propose the EPDF scheduling policy, which prioritizes packets based on a combination of their deadline and a measure of 'debt'—the cumulative deficit in throughput relative to required rates.
• Model the system with a server that introduces a controlled buffering delay to absorb network and traffic variability, improving robustness to burstiness and channel fading.
• Define the capacity region as the set of all achievable (delay, throughput) pairs for each flow, derived from the joint constraints of per-packet delay bounds and per-flow throughput requirements.
• Use a Lyapunov drift analysis to prove that EPDF is capacity-achieving, meaning it supports all strictly feasible (delay, throughput) vectors in the capacity region.
• Implement and test EPDF using real video traces from three movies (Harry Potter, Finding Neverland, Transporter 2) in ns-2 simulations.
• Introduce a modified EPDF variant to improve short-term fairness, analyzing its impact on long-term performance guarantees.

Experimental results

Research questions

• RQ1 What is the precise capacity region for multiple live video streams under per-packet delay and throughput constraints, given buffering at the server?
• RQ2 How can a scheduling policy be designed to achieve the full capacity region without requiring knowledge of future traffic or channel states?
• RQ3 What is the impact of increasing delay bounds on the achievable throughput region for live video streaming?
• RQ4 How does frame size (M) affect the short-term and long-term performance of the EPDF policy?
• RQ5 To what extent can short-term performance be improved without significantly degrading long-term fairness or throughput guarantees?

Key findings

• The EPDF policy outperforms LDF, EDF, and a heuristic from prior work by a large margin in all three video traces, with throughput differences up to 0.13 for the same delay bound.
• Increasing the delay bound from 0.1s to 1s improves the maximum achievable throughput by up to 0.2, and further increasing it to 10s adds another 0.21 in some cases.
• Despite theoretical concerns, modifying EPDF to enhance short-term performance results in negligible impact on long-term performance, especially under non-adversarial traffic.
• The achieved throughput regions for frame sizes M=1 and M=1000 are nearly indistinguishable in long-term performance, with differences less than 0.03 across all traces.
• Smaller frame sizes (M=1) lead to significantly more stable short-term delivery, with less fluctuation in packet counts per second, compared to M=1000.
• The EPDF policy is proven to be capacity-achieving, supporting all strictly feasible (delay, throughput) vectors in the capacity region under the given model.