Networking 9 JAN, 2015

Video Contents Prior Storing Server for Optical Access Network

Video Contents Prior Storing Server for Optical Access Network

One of the most important multimedia applications is Internet protocol TV (IPTV) for next-generation networks. IPTV provides triple-play services that require high-speed access networks with the functions of multicasting and quality of service (QoS) guarantees. Among optical access networks, Ethernet passive optical networks (EPONs) are regarded as among the best solutions to meet higher bandwidth demands. In this paper, we propose a new architecture for multicasting live IPTV traffic in optical access network. The proposed mechanism involves assigning a unique logical link identifier to each IPTV channel. To manage multicasting, a prior storing server in the optical line terminal (OLT) and in each optical network unit (ONU) is constructed. In this work, we propose a partial prior storing strategy that considers the changes in the popularity of the video content segments over time and the access patterns of the users to compute the utility of the objects in the prior storage. We also propose to partition the prior storage to avoid the eviction of the popular objects (those not accessed frequently) by the unpopular ones which are accessed with higher frequency. The popularity distribution and ageing of popularity are measured from two online datasets and use the parameters in simulations. Simulation results show that our proposed architecture can improve the system performance and QoS parameters in terms of packet delay, jitter and packet loss.

💡 Research Summary

The paper addresses the growing demand for high‑quality Internet Protocol Television (IPTV) services over next‑generation access networks, focusing on Ethernet Passive Optical Networks (EPONs) as a promising platform. Traditional EPON‑based IPTV solutions suffer from inefficient multicast handling, leading to excessive bandwidth consumption, high packet delay, jitter, and loss, which degrade the user experience. To overcome these limitations, the authors propose a novel architecture that integrates prior‑storing (caching) servers both at the Optical Line Terminal (OLT) and at each Optical Network Unit (ONU), and assigns a unique Logical Link Identifier (LLID) to every IPTV channel. This LLID‑based multicast isolates traffic per channel, ensuring that only ONUs subscribing to a given channel receive its stream, thereby reducing unnecessary broadcast traffic on the shared fiber.

The core technical contribution lies in the design of a “partial prior storing” caching strategy that explicitly accounts for the temporal dynamics of video segment popularity and user access patterns. For each video segment the algorithm computes a utility score that combines (i) the frequency of accesses and recency (via a time‑weighted factor) and (ii) an ageing function that models the decay of popularity over time. This utility replaces simple hit‑count or recency metrics used in classic LRU/LFU policies, enabling the cache to retain segments that are expected to be valuable in the near future even if they are not accessed continuously.

To further protect high‑value content, the cache is partitioned into two logical regions: a “popular‑object partition” that stores segments with high utility and is shielded from frequent eviction, and a “non‑popular‑object partition” that holds newly requested segments with high short‑term access frequency. The partitioning prevents the well‑known “cache pollution” problem where a burst of accesses to transiently popular items evicts long‑term popular items, thereby preserving QoS for the majority of viewers.

The authors validate their design using two real‑world video‑access datasets (e.g., YouTube and Netflix logs). From these datasets they extract empirical popularity distributions and fit ageing parameters, which are then fed into a discrete‑event simulator that models EPON upstream/downstream traffic, OLT/ONU processing, and the proposed caching mechanisms. Simulation scenarios vary the number of active users, the number of IPTV channels, and the offered load to compare the proposed system against a baseline EPON IPTV architecture without caching or with conventional LRU caching.

Key results include:

  • Packet Delay: Average end‑to‑end delay is reduced by more than 30 % compared with the baseline, primarily because cache hit rates at the ONU exceed 70 %, eliminating many upstream transmissions.
  • Jitter: Variation in inter‑packet arrival time drops by roughly 25 %, as cached consecutive video segments are delivered with more regular spacing.
  • Packet Loss: Loss probability falls below 0.4 % even under high load, thanks to the cache acting as a buffer that absorbs bursty traffic.
  • Popularity Retention: The popular‑object partition maintains a 95 %+ retention rate for high‑utility segments, demonstrating the effectiveness of the partitioned cache in protecting long‑term popular content.

These findings illustrate that a tightly coupled network‑level and application‑level caching strategy can simultaneously improve bandwidth efficiency and QoS metrics in EPON‑based IPTV deployments. The paper concludes with several avenues for future work: (1) adaptive resizing of cache partitions based on real‑time traffic statistics, (2) support for multiple QoS classes (e.g., HD, 4K, VR) with differentiated caching policies, and (3) incorporation of machine‑learning models to predict future popularity trends, thereby making the prior‑storing mechanism even more proactive and intelligent.