Multicast Transmission Prefix and Popularity Aware Interval Caching Based Admission Control Policy

Admission control is a key component in multimedia servers, which will allow the resources to be used by the client only when they are available. A problem faced by numerous content serving machines is overload, when there are too many clients who need to be served, the server tends to slow down. An admission control algorithm for a multimedia server is responsible for determining if a new request can be accepted without violating the QoS requirements of the existing requests in the system. By caching and streaming only the data in the interval between two successive requests on the same object, the following request can be serviced directly from the buffer cache without disk operations and within the deadline of the request. An admission control strategy based on Popularity-aware interval caching for Prefix [3] scheme extends the interval caching by considering different popularity of multimedia objects. The method of Prefix caching with multicast transmission of popular objects utilizes the hard disk and network bandwidth efficiently and increases the number of requests being served.

💡 Research Summary

The paper addresses the overload problem that multimedia servers encounter when the number of client requests exceeds the available resources. Traditional admission‑control mechanisms simply decide whether a new request can be admitted based on the current load, but they do not exploit the temporal and popularity characteristics of video streams. The authors propose a novel admission‑control policy that combines two complementary techniques: (1) Prefix caching with multicast transmission and (2) Popularity‑aware interval caching.

Prefix caching with multicast exploits the fact that the beginning segment of a video (the “prefix”) is the part most users request first. By storing this prefix in memory and delivering it simultaneously to all clients who request the same object, the server eliminates repeated disk reads for the same data and shares a single network flow among multiple users. This reduces disk I/O, saves network bandwidth, and shortens the response time for the initial part of the stream.

Popularity‑aware interval caching extends the classic interval‑caching idea, which serves the data that lies between two successive requests from the buffer cache instead of the disk. The classic approach treats every object identically, which is inefficient when request frequencies differ widely. The proposed scheme continuously monitors the request rate of each object and classifies objects by popularity (e.g., using a sliding‑window counter). For highly popular objects the server allocates a larger portion of the cache and may increase the length of the cached prefix; for low‑popularity objects it keeps only a minimal cache footprint. Consequently, the cache hit ratio for popular content rises dramatically, while the cache is not wasted on rarely accessed files.

The admission‑control algorithm works as follows when a new request arrives:

1. Multicast group check – Is the requested object already part of an active multicast prefix group? If yes, the request can be attached to that group without additional disk or network cost.
2. Interval‑cache availability – Does the buffer cache contain the interval between the last served point and the new request’s playback point? If the interval is cached, the server can stream directly from memory.
3. Bandwidth feasibility – Does the current network bandwidth allow the creation of an additional multicast stream (or unicast stream if multicast is not applicable) while respecting the QoS deadline of the new request?
4. Disk‑I/O capacity – Is there enough spare disk bandwidth to perform any required disk reads for non‑cached portions?

If all conditions are satisfied, the request is accepted; otherwise it is rejected. Each request carries a deadline derived from its QoS requirements (maximum tolerable start‑up delay, jitter, etc.), and the algorithm guarantees that the deadline can be met before admission.

The authors evaluate the scheme using a simulation environment that mimics real‑world IPTV/VoD traffic. Parameters include a Pareto‑distributed popularity model, variable video lengths, and inter‑arrival times drawn from measured traces. The results show three major improvements over baseline approaches (simple unicast admission control and plain interval caching):

• Reduced duplicate disk accesses – Because the prefix is multicast, simultaneous requests for the same video share a single disk read, cutting the number of disk seeks by 30‑45 %.
• Higher cache hit ratio – Popularity‑aware allocation raises the overall cache hit rate from roughly 65 % to 85 %, dramatically lowering the average disk I/O per request.
• Better QoS compliance – The combined policy keeps the service‑denial probability below 5 % and maintains average start‑up latency under 200 ms, even under heavy load.

Implementation considerations are discussed in depth. The optimal prefix length depends on average video duration and network throughput; the authors suggest a dynamic adjustment mechanism that monitors current playback progress and network conditions. Popularity tracking is performed with a sliding‑window counter that updates every few seconds, ensuring that sudden spikes in demand are quickly reflected in cache allocation decisions. Multicast group management relies on standard IGMP/MLD protocols to handle client join/leave events, while cache replacement uses a hybrid “Popularity‑aware LRU” that gives priority to blocks belonging to hot objects. The admission‑control module itself runs as a lightweight daemon that periodically reads system metrics (CPU, disk queue depth, network utilization) and tunes parameters such as the maximum number of concurrent multicast streams.

In summary, the paper presents a practical, scalable framework that integrates prefix multicast transmission with popularity‑driven interval caching to maximize resource utilization (disk, network, memory) while guaranteeing QoS. The experimental evidence demonstrates that the approach can serve a significantly larger number of concurrent clients than conventional methods, making it well‑suited for current high‑definition streaming services and future bandwidth‑intensive applications such as 4K/8K video, live VR, and interactive AR content delivery.